drivers/net/ethernet/ibm/ibmvnic.c

   1 /**************************************************************************/
   2 /*                                                                        */
   3 /*  IBM System i and System p Virtual NIC Device Driver                   */
   4 /*  Copyright (C) 2014 IBM Corp.                                          */
   5 /*  Santiago Leon (santi_leon@yahoo.com)                                  */
   6 /*  Thomas Falcon (tlfalcon@linux.vnet.ibm.com)                           */
   7 /*  John Allen (jallen@linux.vnet.ibm.com)                                */
   8 /*                                                                        */
   9 /*  This program is free software; you can redistribute it and/or modify  */
  10 /*  it under the terms of the GNU General Public License as published by  */
  11 /*  the Free Software Foundation; either version 2 of the License, or     */
  12 /*  (at your option) any later version.                                   */
  13 /*                                                                        */
  14 /*  This program is distributed in the hope that it will be useful,       */
  15 /*  but WITHOUT ANY WARRANTY; without even the implied warranty of        */
  16 /*  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the         */
  17 /*  GNU General Public License for more details.                          */
  18 /*                                                                        */
  19 /*  You should have received a copy of the GNU General Public License     */
  20 /*  along with this program.                                              */
  21 /*                                                                        */
  22 /* This module contains the implementation of a virtual ethernet device   */
  23 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN    */
  24 /* option of the RS/6000 Platform Architecture to interface with virtual  */
  25 /* ethernet NICs that are presented to the partition by the hypervisor.   */
  26 /*                                                                         */
  27 /* Messages are passed between the VNIC driver and the VNIC server using  */
  28 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to  */
  29 /* issue and receive commands that initiate communication with the server */
  30 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but    */
  31 /* are used by the driver to notify the server that a packet is           */
  32 /* ready for transmission or that a buffer has been added to receive a    */
  33 /* packet. Subsequently, sCRQs are used by the server to notify the       */
  34 /* driver that a packet transmission has been completed or that a packet  */
  35 /* has been received and placed in a waiting buffer.                      */
  36 /*                                                                        */
  37 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in    */
  38 /* which skbs are DMA mapped and immediately unmapped when the transmit   */
  39 /* or receive has been completed, the VNIC driver is required to use      */
  40 /* "long term mapping". This entails that large, continuous DMA mapped    */
  41 /* buffers are allocated on driver initialization and these buffers are   */
  42 /* then continuously reused to pass skbs to and from the VNIC server.     */
  43 /*                                                                        */
  44 /**************************************************************************/
  45
  46 #include <linux/module.h>
  47 #include <linux/moduleparam.h>
  48 #include <linux/types.h>
  49 #include <linux/errno.h>
  50 #include <linux/completion.h>
  51 #include <linux/ioport.h>
  52 #include <linux/dma-mapping.h>
  53 #include <linux/kernel.h>
  54 #include <linux/netdevice.h>
  55 #include <linux/etherdevice.h>
  56 #include <linux/skbuff.h>
  57 #include <linux/init.h>
  58 #include <linux/delay.h>
  59 #include <linux/mm.h>
  60 #include <linux/ethtool.h>
  61 #include <linux/proc_fs.h>
  62 #include <linux/if_arp.h>
  63 #include <linux/in.h>
  64 #include <linux/ip.h>
  65 #include <linux/ipv6.h>
  66 #include <linux/irq.h>
  67 #include <linux/kthread.h>
  68 #include <linux/seq_file.h>
  69 #include <linux/interrupt.h>
  70 #include <net/net_namespace.h>
  71 #include <asm/hvcall.h>
  72 #include <linux/atomic.h>
  73 #include <asm/vio.h>
  74 #include <asm/iommu.h>
  75 #include <linux/uaccess.h>
  76 #include <asm/firmware.h>
  77 #include <linux/workqueue.h>
  78 #include <linux/if_vlan.h>
  79 #include <linux/utsname.h>
  80
  81 #include "ibmvnic.h"
  82
  83 static const char ibmvnic_driver_name[] = "ibmvnic";
  84 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
  85
  86 MODULE_AUTHOR("Santiago Leon");
  87 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
  88 MODULE_LICENSE("GPL");
  89 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
  90
  91 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
  92 static int ibmvnic_remove(struct vio_dev *);
  93 static void release_sub_crqs(struct ibmvnic_adapter *);
  94 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
  95 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
  96 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
  97 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
  98 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
  99                        union sub_crq *sub_crq);
 100 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
 101 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
 102 static int enable_scrq_irq(struct ibmvnic_adapter *,
 103                            struct ibmvnic_sub_crq_queue *);
 104 static int disable_scrq_irq(struct ibmvnic_adapter *,
 105                             struct ibmvnic_sub_crq_queue *);
 106 static int pending_scrq(struct ibmvnic_adapter *,
 107                         struct ibmvnic_sub_crq_queue *);
 108 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
 109                                         struct ibmvnic_sub_crq_queue *);
 110 static int ibmvnic_poll(struct napi_struct *napi, int data);
 111 static void send_map_query(struct ibmvnic_adapter *adapter);
 112 static void send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
 113 static void send_request_unmap(struct ibmvnic_adapter *, u8);
 114 static void send_login(struct ibmvnic_adapter *adapter);
 115 static void send_cap_queries(struct ibmvnic_adapter *adapter);
 116 static int init_sub_crqs(struct ibmvnic_adapter *);
 117 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
 118 static int ibmvnic_init(struct ibmvnic_adapter *);
 119 static void release_crq_queue(struct ibmvnic_adapter *);
 120 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p);
 121
 122 struct ibmvnic_stat {
 123         char name[ETH_GSTRING_LEN];
 124         int offset;
 125 };
 126
 127 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
 128                              offsetof(struct ibmvnic_statistics, stat))
 129 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
 130
 131 static const struct ibmvnic_stat ibmvnic_stats[] = {
 132         {"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
 133         {"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
 134         {"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
 135         {"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
 136         {"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
 137         {"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
 138         {"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
 139         {"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
 140         {"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
 141         {"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
 142         {"align_errors", IBMVNIC_STAT_OFF(align_errors)},
 143         {"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
 144         {"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
 145         {"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
 146         {"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
 147         {"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
 148         {"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
 149         {"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
 150         {"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
 151         {"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
 152         {"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
 153         {"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
 154 };
 155
 156 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
 157                           unsigned long length, unsigned long *number,
 158                           unsigned long *irq)
 159 {
 160         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
 161         long rc;
 162
 163         rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
 164         *number = retbuf[0];
 165         *irq = retbuf[1];
 166
 167         return rc;
 168 }
 169
 170 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
 171                                 struct ibmvnic_long_term_buff *ltb, int size)
 172 {
 173         struct device *dev = &adapter->vdev->dev;
 174
 175         ltb->size = size;
 176         ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
 177                                        GFP_KERNEL);
 178
 179         if (!ltb->buff) {
 180                 dev_err(dev, "Couldn't alloc long term buffer\n");
 181                 return -ENOMEM;
 182         }
 183         ltb->map_id = adapter->map_id;
 184         adapter->map_id++;
 185
 186         init_completion(&adapter->fw_done);
 187         send_request_map(adapter, ltb->addr,
 188                          ltb->size, ltb->map_id);
 189         wait_for_completion(&adapter->fw_done);
 190
 191         if (adapter->fw_done_rc) {
 192                 dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
 193                         adapter->fw_done_rc);
 194                 return -1;
 195         }
 196         return 0;
 197 }
 198
 199 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
 200                                 struct ibmvnic_long_term_buff *ltb)
 201 {
 202         struct device *dev = &adapter->vdev->dev;
 203
 204         if (!ltb->buff)
 205                 return;
 206
 207         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
 208             adapter->reset_reason != VNIC_RESET_MOBILITY)
 209                 send_request_unmap(adapter, ltb->map_id);
 210         dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 211 }
 212
 213 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
 214                                 struct ibmvnic_long_term_buff *ltb)
 215 {
 216         memset(ltb->buff, 0, ltb->size);
 217
 218         init_completion(&adapter->fw_done);
 219         send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
 220         wait_for_completion(&adapter->fw_done);
 221
 222         if (adapter->fw_done_rc) {
 223                 dev_info(&adapter->vdev->dev,
 224                          "Reset failed, attempting to free and reallocate buffer\n");
 225                 free_long_term_buff(adapter, ltb);
 226                 return alloc_long_term_buff(adapter, ltb, ltb->size);
 227         }
 228         return 0;
 229 }
 230
 231 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
 232 {
 233         int i;
 234
 235         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 236              i++)
 237                 adapter->rx_pool[i].active = 0;
 238 }
 239
 240 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
 241                               struct ibmvnic_rx_pool *pool)
 242 {
 243         int count = pool->size - atomic_read(&pool->available);
 244         struct device *dev = &adapter->vdev->dev;
 245         int buffers_added = 0;
 246         unsigned long lpar_rc;
 247         union sub_crq sub_crq;
 248         struct sk_buff *skb;
 249         unsigned int offset;
 250         dma_addr_t dma_addr;
 251         unsigned char *dst;
 252         u64 *handle_array;
 253         int shift = 0;
 254         int index;
 255         int i;
 256
 257         if (!pool->active)
 258                 return;
 259
 260         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 261                                       be32_to_cpu(adapter->login_rsp_buf->
 262                                       off_rxadd_subcrqs));
 263
 264         for (i = 0; i < count; ++i) {
 265                 skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
 266                 if (!skb) {
 267                         dev_err(dev, "Couldn't replenish rx buff\n");
 268                         adapter->replenish_no_mem++;
 269                         break;
 270                 }
 271
 272                 index = pool->free_map[pool->next_free];
 273
 274                 if (pool->rx_buff[index].skb)
 275                         dev_err(dev, "Inconsistent free_map!\n");
 276
 277                 /* Copy the skb to the long term mapped DMA buffer */
 278                 offset = index * pool->buff_size;
 279                 dst = pool->long_term_buff.buff + offset;
 280                 memset(dst, 0, pool->buff_size);
 281                 dma_addr = pool->long_term_buff.addr + offset;
 282                 pool->rx_buff[index].data = dst;
 283
 284                 pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
 285                 pool->rx_buff[index].dma = dma_addr;
 286                 pool->rx_buff[index].skb = skb;
 287                 pool->rx_buff[index].pool_index = pool->index;
 288                 pool->rx_buff[index].size = pool->buff_size;
 289
 290                 memset(&sub_crq, 0, sizeof(sub_crq));
 291                 sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
 292                 sub_crq.rx_add.correlator =
 293                     cpu_to_be64((u64)&pool->rx_buff[index]);
 294                 sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
 295                 sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
 296
 297                 /* The length field of the sCRQ is defined to be 24 bits so the
 298                  * buffer size needs to be left shifted by a byte before it is
 299                  * converted to big endian to prevent the last byte from being
 300                  * truncated.
 301                  */
 302 #ifdef __LITTLE_ENDIAN__
 303                 shift = 8;
 304 #endif
 305                 sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
 306
 307                 lpar_rc = send_subcrq(adapter, handle_array[pool->index],
 308                                       &sub_crq);
 309                 if (lpar_rc != H_SUCCESS)
 310                         goto failure;
 311
 312                 buffers_added++;
 313                 adapter->replenish_add_buff_success++;
 314                 pool->next_free = (pool->next_free + 1) % pool->size;
 315         }
 316         atomic_add(buffers_added, &pool->available);
 317         return;
 318
 319 failure:
 320         dev_info(dev, "replenish pools failure\n");
 321         pool->free_map[pool->next_free] = index;
 322         pool->rx_buff[index].skb = NULL;
 323         if (!dma_mapping_error(dev, dma_addr))
 324                 dma_unmap_single(dev, dma_addr, pool->buff_size,
 325                                  DMA_FROM_DEVICE);
 326
 327         dev_kfree_skb_any(skb);
 328         adapter->replenish_add_buff_failure++;
 329         atomic_add(buffers_added, &pool->available);
 330
 331         if (lpar_rc == H_CLOSED) {
 332                 /* Disable buffer pool replenishment and report carrier off if
 333                  * queue is closed. Firmware guarantees that a signal will
 334                  * be sent to the driver, triggering a reset.
 335                  */
 336                 deactivate_rx_pools(adapter);
 337                 netif_carrier_off(adapter->netdev);
 338         }
 339 }
 340
 341 static void replenish_pools(struct ibmvnic_adapter *adapter)
 342 {
 343         int i;
 344
 345         adapter->replenish_task_cycles++;
 346         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 347              i++) {
 348                 if (adapter->rx_pool[i].active)
 349                         replenish_rx_pool(adapter, &adapter->rx_pool[i]);
 350         }
 351 }
 352
 353 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
 354 {
 355         kfree(adapter->tx_stats_buffers);
 356         kfree(adapter->rx_stats_buffers);
 357 }
 358
 359 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
 360 {
 361         adapter->tx_stats_buffers =
 362                                 kcalloc(adapter->req_tx_queues,
 363                                         sizeof(struct ibmvnic_tx_queue_stats),
 364                                         GFP_KERNEL);
 365         if (!adapter->tx_stats_buffers)
 366                 return -ENOMEM;
 367
 368         adapter->rx_stats_buffers =
 369                                 kcalloc(adapter->req_rx_queues,
 370                                         sizeof(struct ibmvnic_rx_queue_stats),
 371                                         GFP_KERNEL);
 372         if (!adapter->rx_stats_buffers)
 373                 return -ENOMEM;
 374
 375         return 0;
 376 }
 377
 378 static void release_stats_token(struct ibmvnic_adapter *adapter)
 379 {
 380         struct device *dev = &adapter->vdev->dev;
 381
 382         if (!adapter->stats_token)
 383                 return;
 384
 385         dma_unmap_single(dev, adapter->stats_token,
 386                          sizeof(struct ibmvnic_statistics),
 387                          DMA_FROM_DEVICE);
 388         adapter->stats_token = 0;
 389 }
 390
 391 static int init_stats_token(struct ibmvnic_adapter *adapter)
 392 {
 393         struct device *dev = &adapter->vdev->dev;
 394         dma_addr_t stok;
 395
 396         stok = dma_map_single(dev, &adapter->stats,
 397                               sizeof(struct ibmvnic_statistics),
 398                               DMA_FROM_DEVICE);
 399         if (dma_mapping_error(dev, stok)) {
 400                 dev_err(dev, "Couldn't map stats buffer\n");
 401                 return -1;
 402         }
 403
 404         adapter->stats_token = stok;
 405         netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
 406         return 0;
 407 }
 408
 409 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
 410 {
 411         struct ibmvnic_rx_pool *rx_pool;
 412         int rx_scrqs;
 413         int i, j, rc;
 414         u64 *size_array;
 415
 416         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 417                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 418
 419         rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 420         for (i = 0; i < rx_scrqs; i++) {
 421                 rx_pool = &adapter->rx_pool[i];
 422
 423                 netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
 424
 425                 if (rx_pool->buff_size != be64_to_cpu(size_array[i])) {
 426                         free_long_term_buff(adapter, &rx_pool->long_term_buff);
 427                         rx_pool->buff_size = be64_to_cpu(size_array[i]);
 428                         alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 429                                              rx_pool->size *
 430                                              rx_pool->buff_size);
 431                 } else {
 432                         rc = reset_long_term_buff(adapter,
 433                                                   &rx_pool->long_term_buff);
 434                 }
 435
 436                 if (rc)
 437                         return rc;
 438
 439                 for (j = 0; j < rx_pool->size; j++)
 440                         rx_pool->free_map[j] = j;
 441
 442                 memset(rx_pool->rx_buff, 0,
 443                        rx_pool->size * sizeof(struct ibmvnic_rx_buff));
 444
 445                 atomic_set(&rx_pool->available, 0);
 446                 rx_pool->next_alloc = 0;
 447                 rx_pool->next_free = 0;
 448                 rx_pool->active = 1;
 449         }
 450
 451         return 0;
 452 }
 453
 454 static void release_rx_pools(struct ibmvnic_adapter *adapter)
 455 {
 456         struct ibmvnic_rx_pool *rx_pool;
 457         int i, j;
 458
 459         if (!adapter->rx_pool)
 460                 return;
 461
 462         for (i = 0; i < adapter->num_active_rx_pools; i++) {
 463                 rx_pool = &adapter->rx_pool[i];
 464
 465                 netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
 466
 467                 kfree(rx_pool->free_map);
 468                 free_long_term_buff(adapter, &rx_pool->long_term_buff);
 469
 470                 if (!rx_pool->rx_buff)
 471                         continue;
 472
 473                 for (j = 0; j < rx_pool->size; j++) {
 474                         if (rx_pool->rx_buff[j].skb) {
 475                                 dev_kfree_skb_any(rx_pool->rx_buff[i].skb);
 476                                 rx_pool->rx_buff[i].skb = NULL;
 477                         }
 478                 }
 479
 480                 kfree(rx_pool->rx_buff);
 481         }
 482
 483         kfree(adapter->rx_pool);
 484         adapter->rx_pool = NULL;
 485         adapter->num_active_rx_pools = 0;
 486 }
 487
 488 static int init_rx_pools(struct net_device *netdev)
 489 {
 490         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 491         struct device *dev = &adapter->vdev->dev;
 492         struct ibmvnic_rx_pool *rx_pool;
 493         int rxadd_subcrqs;
 494         u64 *size_array;
 495         int i, j;
 496
 497         rxadd_subcrqs =
 498                 be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 499         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 500                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 501
 502         adapter->rx_pool = kcalloc(rxadd_subcrqs,
 503                                    sizeof(struct ibmvnic_rx_pool),
 504                                    GFP_KERNEL);
 505         if (!adapter->rx_pool) {
 506                 dev_err(dev, "Failed to allocate rx pools\n");
 507                 return -1;
 508         }
 509
 510         adapter->num_active_rx_pools = 0;
 511
 512         for (i = 0; i < rxadd_subcrqs; i++) {
 513                 rx_pool = &adapter->rx_pool[i];
 514
 515                 netdev_dbg(adapter->netdev,
 516                            "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
 517                            i, adapter->req_rx_add_entries_per_subcrq,
 518                            be64_to_cpu(size_array[i]));
 519
 520                 rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
 521                 rx_pool->index = i;
 522                 rx_pool->buff_size = be64_to_cpu(size_array[i]);
 523                 rx_pool->active = 1;
 524
 525                 rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
 526                                             GFP_KERNEL);
 527                 if (!rx_pool->free_map) {
 528                         release_rx_pools(adapter);
 529                         return -1;
 530                 }
 531
 532                 rx_pool->rx_buff = kcalloc(rx_pool->size,
 533                                            sizeof(struct ibmvnic_rx_buff),
 534                                            GFP_KERNEL);
 535                 if (!rx_pool->rx_buff) {
 536                         dev_err(dev, "Couldn't alloc rx buffers\n");
 537                         release_rx_pools(adapter);
 538                         return -1;
 539                 }
 540
 541                 if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 542                                          rx_pool->size * rx_pool->buff_size)) {
 543                         release_rx_pools(adapter);
 544                         return -1;
 545                 }
 546
 547                 for (j = 0; j < rx_pool->size; ++j)
 548                         rx_pool->free_map[j] = j;
 549
 550                 atomic_set(&rx_pool->available, 0);
 551                 rx_pool->next_alloc = 0;
 552                 rx_pool->next_free = 0;
 553         }
 554
 555         adapter->num_active_rx_pools = rxadd_subcrqs;
 556
 557         return 0;
 558 }
 559
 560 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
 561 {
 562         struct ibmvnic_tx_pool *tx_pool;
 563         int tx_scrqs;
 564         int i, j, rc;
 565
 566         tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 567         for (i = 0; i < tx_scrqs; i++) {
 568                 netdev_dbg(adapter->netdev, "Re-setting tx_pool[%d]\n", i);
 569
 570                 tx_pool = &adapter->tx_pool[i];
 571
 572                 rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
 573                 if (rc)
 574                         return rc;
 575
 576                 rc = reset_long_term_buff(adapter, &tx_pool->tso_ltb);
 577                 if (rc)
 578                         return rc;
 579
 580                 memset(tx_pool->tx_buff, 0,
 581                        adapter->req_tx_entries_per_subcrq *
 582                        sizeof(struct ibmvnic_tx_buff));
 583
 584                 for (j = 0; j < adapter->req_tx_entries_per_subcrq; j++)
 585                         tx_pool->free_map[j] = j;
 586
 587                 tx_pool->consumer_index = 0;
 588                 tx_pool->producer_index = 0;
 589                 tx_pool->tso_index = 0;
 590         }
 591
 592         return 0;
 593 }
 594
 595 static void release_vpd_data(struct ibmvnic_adapter *adapter)
 596 {
 597         if (!adapter->vpd)
 598                 return;
 599
 600         kfree(adapter->vpd->buff);
 601         kfree(adapter->vpd);
 602 }
 603
 604 static void release_tx_pools(struct ibmvnic_adapter *adapter)
 605 {
 606         struct ibmvnic_tx_pool *tx_pool;
 607         int i;
 608
 609         if (!adapter->tx_pool)
 610                 return;
 611
 612         for (i = 0; i < adapter->num_active_tx_pools; i++) {
 613                 netdev_dbg(adapter->netdev, "Releasing tx_pool[%d]\n", i);
 614                 tx_pool = &adapter->tx_pool[i];
 615                 kfree(tx_pool->tx_buff);
 616                 free_long_term_buff(adapter, &tx_pool->long_term_buff);
 617                 free_long_term_buff(adapter, &tx_pool->tso_ltb);
 618                 kfree(tx_pool->free_map);
 619         }
 620
 621         kfree(adapter->tx_pool);
 622         adapter->tx_pool = NULL;
 623         adapter->num_active_tx_pools = 0;
 624 }
 625
 626 static int init_tx_pools(struct net_device *netdev)
 627 {
 628         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 629         struct device *dev = &adapter->vdev->dev;
 630         struct ibmvnic_tx_pool *tx_pool;
 631         int tx_subcrqs;
 632         int i, j;
 633
 634         tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 635         adapter->tx_pool = kcalloc(tx_subcrqs,
 636                                    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 637         if (!adapter->tx_pool)
 638                 return -1;
 639
 640         adapter->num_active_tx_pools = 0;
 641
 642         for (i = 0; i < tx_subcrqs; i++) {
 643                 tx_pool = &adapter->tx_pool[i];
 644
 645                 netdev_dbg(adapter->netdev,
 646                            "Initializing tx_pool[%d], %lld buffs\n",
 647                            i, adapter->req_tx_entries_per_subcrq);
 648
 649                 tx_pool->tx_buff = kcalloc(adapter->req_tx_entries_per_subcrq,
 650                                            sizeof(struct ibmvnic_tx_buff),
 651                                            GFP_KERNEL);
 652                 if (!tx_pool->tx_buff) {
 653                         dev_err(dev, "tx pool buffer allocation failed\n");
 654                         release_tx_pools(adapter);
 655                         return -1;
 656                 }
 657
 658                 if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
 659                                          adapter->req_tx_entries_per_subcrq *
 660                                          adapter->req_mtu)) {
 661                         release_tx_pools(adapter);
 662                         return -1;
 663                 }
 664
 665                 /* alloc TSO ltb */
 666                 if (alloc_long_term_buff(adapter, &tx_pool->tso_ltb,
 667                                          IBMVNIC_TSO_BUFS *
 668                                          IBMVNIC_TSO_BUF_SZ)) {
 669                         release_tx_pools(adapter);
 670                         return -1;
 671                 }
 672
 673                 tx_pool->tso_index = 0;
 674
 675                 tx_pool->free_map = kcalloc(adapter->req_tx_entries_per_subcrq,
 676                                             sizeof(int), GFP_KERNEL);
 677                 if (!tx_pool->free_map) {
 678                         release_tx_pools(adapter);
 679                         return -1;
 680                 }
 681
 682                 for (j = 0; j < adapter->req_tx_entries_per_subcrq; j++)
 683                         tx_pool->free_map[j] = j;
 684
 685                 tx_pool->consumer_index = 0;
 686                 tx_pool->producer_index = 0;
 687         }
 688
 689         adapter->num_active_tx_pools = tx_subcrqs;
 690
 691         return 0;
 692 }
 693
 694 static void release_error_buffers(struct ibmvnic_adapter *adapter)
 695 {
 696         struct device *dev = &adapter->vdev->dev;
 697         struct ibmvnic_error_buff *error_buff, *tmp;
 698         unsigned long flags;
 699
 700         spin_lock_irqsave(&adapter->error_list_lock, flags);
 701         list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list) {
 702                 list_del(&error_buff->list);
 703                 dma_unmap_single(dev, error_buff->dma, error_buff->len,
 704                                  DMA_FROM_DEVICE);
 705                 kfree(error_buff->buff);
 706                 kfree(error_buff);
 707         }
 708         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
 709 }
 710
 711 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
 712 {
 713         int i;
 714
 715         if (adapter->napi_enabled)
 716                 return;
 717
 718         for (i = 0; i < adapter->req_rx_queues; i++)
 719                 napi_enable(&adapter->napi[i]);
 720
 721         adapter->napi_enabled = true;
 722 }
 723
 724 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
 725 {
 726         int i;
 727
 728         if (!adapter->napi_enabled)
 729                 return;
 730
 731         for (i = 0; i < adapter->req_rx_queues; i++) {
 732                 netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
 733                 napi_disable(&adapter->napi[i]);
 734         }
 735
 736         adapter->napi_enabled = false;
 737 }
 738
 739 static int ibmvnic_login(struct net_device *netdev)
 740 {
 741         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 742         unsigned long timeout = msecs_to_jiffies(30000);
 743         struct device *dev = &adapter->vdev->dev;
 744         int rc;
 745
 746         do {
 747                 if (adapter->renegotiate) {
 748                         adapter->renegotiate = false;
 749                         release_sub_crqs(adapter);
 750
 751                         reinit_completion(&adapter->init_done);
 752                         send_cap_queries(adapter);
 753                         if (!wait_for_completion_timeout(&adapter->init_done,
 754                                                          timeout)) {
 755                                 dev_err(dev, "Capabilities query timeout\n");
 756                                 return -1;
 757                         }
 758                         rc = init_sub_crqs(adapter);
 759                         if (rc) {
 760                                 dev_err(dev,
 761                                         "Initialization of SCRQ's failed\n");
 762                                 return -1;
 763                         }
 764                         rc = init_sub_crq_irqs(adapter);
 765                         if (rc) {
 766                                 dev_err(dev,
 767                                         "Initialization of SCRQ's irqs failed\n");
 768                                 return -1;
 769                         }
 770                 }
 771
 772                 reinit_completion(&adapter->init_done);
 773                 send_login(adapter);
 774                 if (!wait_for_completion_timeout(&adapter->init_done,
 775                                                  timeout)) {
 776                         dev_err(dev, "Login timeout\n");
 777                         return -1;
 778                 }
 779         } while (adapter->renegotiate);
 780
 781         /* handle pending MAC address changes after successful login */
 782         if (adapter->mac_change_pending) {
 783                 __ibmvnic_set_mac(netdev, &adapter->desired.mac);
 784                 adapter->mac_change_pending = false;
 785         }
 786
 787         return 0;
 788 }
 789
 790 static void release_resources(struct ibmvnic_adapter *adapter)
 791 {
 792         int i;
 793
 794         release_vpd_data(adapter);
 795
 796         release_tx_pools(adapter);
 797         release_rx_pools(adapter);
 798
 799         release_stats_token(adapter);
 800         release_stats_buffers(adapter);
 801         release_error_buffers(adapter);
 802
 803         if (adapter->napi) {
 804                 for (i = 0; i < adapter->req_rx_queues; i++) {
 805                         if (&adapter->napi[i]) {
 806                                 netdev_dbg(adapter->netdev,
 807                                            "Releasing napi[%d]\n", i);
 808                                 netif_napi_del(&adapter->napi[i]);
 809                         }
 810                 }
 811         }
 812 }
 813
 814 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
 815 {
 816         struct net_device *netdev = adapter->netdev;
 817         unsigned long timeout = msecs_to_jiffies(30000);
 818         union ibmvnic_crq crq;
 819         bool resend;
 820         int rc;
 821
 822         netdev_dbg(netdev, "setting link state %d\n", link_state);
 823
 824         memset(&crq, 0, sizeof(crq));
 825         crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
 826         crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
 827         crq.logical_link_state.link_state = link_state;
 828
 829         do {
 830                 resend = false;
 831
 832                 reinit_completion(&adapter->init_done);
 833                 rc = ibmvnic_send_crq(adapter, &crq);
 834                 if (rc) {
 835                         netdev_err(netdev, "Failed to set link state\n");
 836                         return rc;
 837                 }
 838
 839                 if (!wait_for_completion_timeout(&adapter->init_done,
 840                                                  timeout)) {
 841                         netdev_err(netdev, "timeout setting link state\n");
 842                         return -1;
 843                 }
 844
 845                 if (adapter->init_done_rc == 1) {
 846                         /* Partuial success, delay and re-send */
 847                         mdelay(1000);
 848                         resend = true;
 849                 }
 850         } while (resend);
 851
 852         return 0;
 853 }
 854
 855 static int set_real_num_queues(struct net_device *netdev)
 856 {
 857         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 858         int rc;
 859
 860         netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
 861                    adapter->req_tx_queues, adapter->req_rx_queues);
 862
 863         rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
 864         if (rc) {
 865                 netdev_err(netdev, "failed to set the number of tx queues\n");
 866                 return rc;
 867         }
 868
 869         rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
 870         if (rc)
 871                 netdev_err(netdev, "failed to set the number of rx queues\n");
 872
 873         return rc;
 874 }
 875
 876 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
 877 {
 878         struct device *dev = &adapter->vdev->dev;
 879         union ibmvnic_crq crq;
 880         int len = 0;
 881
 882         if (adapter->vpd->buff)
 883                 len = adapter->vpd->len;
 884
 885         init_completion(&adapter->fw_done);
 886         crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
 887         crq.get_vpd_size.cmd = GET_VPD_SIZE;
 888         ibmvnic_send_crq(adapter, &crq);
 889         wait_for_completion(&adapter->fw_done);
 890
 891         if (!adapter->vpd->len)
 892                 return -ENODATA;
 893
 894         if (!adapter->vpd->buff)
 895                 adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
 896         else if (adapter->vpd->len != len)
 897                 adapter->vpd->buff =
 898                         krealloc(adapter->vpd->buff,
 899                                  adapter->vpd->len, GFP_KERNEL);
 900
 901         if (!adapter->vpd->buff) {
 902                 dev_err(dev, "Could allocate VPD buffer\n");
 903                 return -ENOMEM;
 904         }
 905
 906         adapter->vpd->dma_addr =
 907                 dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
 908                                DMA_FROM_DEVICE);
 909         if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
 910                 dev_err(dev, "Could not map VPD buffer\n");
 911                 kfree(adapter->vpd->buff);
 912                 return -ENOMEM;
 913         }
 914
 915         reinit_completion(&adapter->fw_done);
 916         crq.get_vpd.first = IBMVNIC_CRQ_CMD;
 917         crq.get_vpd.cmd = GET_VPD;
 918         crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
 919         crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
 920         ibmvnic_send_crq(adapter, &crq);
 921         wait_for_completion(&adapter->fw_done);
 922
 923         return 0;
 924 }
 925
 926 static int init_resources(struct ibmvnic_adapter *adapter)
 927 {
 928         struct net_device *netdev = adapter->netdev;
 929         int i, rc;
 930
 931         rc = set_real_num_queues(netdev);
 932         if (rc)
 933                 return rc;
 934
 935         rc = init_stats_buffers(adapter);
 936         if (rc)
 937                 return rc;
 938
 939         rc = init_stats_token(adapter);
 940         if (rc)
 941                 return rc;
 942
 943         adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
 944         if (!adapter->vpd)
 945                 return -ENOMEM;
 946
 947         /* Vital Product Data (VPD) */
 948         rc = ibmvnic_get_vpd(adapter);
 949         if (rc) {
 950                 netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
 951                 return rc;
 952         }
 953
 954         adapter->map_id = 1;
 955         adapter->napi = kcalloc(adapter->req_rx_queues,
 956                                 sizeof(struct napi_struct), GFP_KERNEL);
 957         if (!adapter->napi)
 958                 return -ENOMEM;
 959
 960         for (i = 0; i < adapter->req_rx_queues; i++) {
 961                 netdev_dbg(netdev, "Adding napi[%d]\n", i);
 962                 netif_napi_add(netdev, &adapter->napi[i], ibmvnic_poll,
 963                                NAPI_POLL_WEIGHT);
 964         }
 965
 966         send_map_query(adapter);
 967
 968         rc = init_rx_pools(netdev);
 969         if (rc)
 970                 return rc;
 971
 972         rc = init_tx_pools(netdev);
 973         return rc;
 974 }
 975
 976 static int __ibmvnic_open(struct net_device *netdev)
 977 {
 978         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 979         enum vnic_state prev_state = adapter->state;
 980         int i, rc;
 981
 982         adapter->state = VNIC_OPENING;
 983         replenish_pools(adapter);
 984         ibmvnic_napi_enable(adapter);
 985
 986         /* We're ready to receive frames, enable the sub-crq interrupts and
 987          * set the logical link state to up
 988          */
 989         for (i = 0; i < adapter->req_rx_queues; i++) {
 990                 netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
 991                 if (prev_state == VNIC_CLOSED)
 992                         enable_irq(adapter->rx_scrq[i]->irq);
 993                 else
 994                         enable_scrq_irq(adapter, adapter->rx_scrq[i]);
 995         }
 996
 997         for (i = 0; i < adapter->req_tx_queues; i++) {
 998                 netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
 999                 if (prev_state == VNIC_CLOSED)
1000                         enable_irq(adapter->tx_scrq[i]->irq);
1001                 else
1002                         enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1003         }
1004
1005         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1006         if (rc) {
1007                 for (i = 0; i < adapter->req_rx_queues; i++)
1008                         napi_disable(&adapter->napi[i]);
1009                 release_resources(adapter);
1010                 return rc;
1011         }
1012
1013         netif_tx_start_all_queues(netdev);
1014
1015         if (prev_state == VNIC_CLOSED) {
1016                 for (i = 0; i < adapter->req_rx_queues; i++)
1017                         napi_schedule(&adapter->napi[i]);
1018         }
1019
1020         adapter->state = VNIC_OPEN;
1021         return rc;
1022 }
1023
1024 static int ibmvnic_open(struct net_device *netdev)
1025 {
1026         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1027         int rc;
1028
1029         mutex_lock(&adapter->reset_lock);
1030
1031         if (adapter->state != VNIC_CLOSED) {
1032                 rc = ibmvnic_login(netdev);
1033                 if (rc) {
1034                         mutex_unlock(&adapter->reset_lock);
1035                         return rc;
1036                 }
1037
1038                 rc = init_resources(adapter);
1039                 if (rc) {
1040                         netdev_err(netdev, "failed to initialize resources\n");
1041                         release_resources(adapter);
1042                         mutex_unlock(&adapter->reset_lock);
1043                         return rc;
1044                 }
1045         }
1046
1047         rc = __ibmvnic_open(netdev);
1048         netif_carrier_on(netdev);
1049
1050         mutex_unlock(&adapter->reset_lock);
1051
1052         return rc;
1053 }
1054
1055 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1056 {
1057         struct ibmvnic_tx_pool *tx_pool;
1058         u64 tx_entries;
1059         int tx_scrqs;
1060         int i, j;
1061
1062         if (!adapter->tx_pool)
1063                 return;
1064
1065         tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
1066         tx_entries = adapter->req_tx_entries_per_subcrq;
1067
1068         /* Free any remaining skbs in the tx buffer pools */
1069         for (i = 0; i < tx_scrqs; i++) {
1070                 tx_pool = &adapter->tx_pool[i];
1071                 if (!tx_pool)
1072                         continue;
1073
1074                 netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1075                 for (j = 0; j < tx_entries; j++) {
1076                         if (tx_pool->tx_buff[j].skb) {
1077                                 dev_kfree_skb_any(tx_pool->tx_buff[j].skb);
1078                                 tx_pool->tx_buff[j].skb = NULL;
1079                         }
1080                 }
1081         }
1082 }
1083
1084 static int __ibmvnic_close(struct net_device *netdev)
1085 {
1086         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1087         int rc = 0;
1088         int i;
1089
1090         adapter->state = VNIC_CLOSING;
1091
1092         /* ensure that transmissions are stopped if called by do_reset */
1093         if (adapter->resetting)
1094                 netif_tx_disable(netdev);
1095         else
1096                 netif_tx_stop_all_queues(netdev);
1097
1098         ibmvnic_napi_disable(adapter);
1099
1100         if (adapter->tx_scrq) {
1101                 for (i = 0; i < adapter->req_tx_queues; i++)
1102                         if (adapter->tx_scrq[i]->irq) {
1103                                 netdev_dbg(adapter->netdev,
1104                                            "Disabling tx_scrq[%d] irq\n", i);
1105                                 disable_irq(adapter->tx_scrq[i]->irq);
1106                         }
1107         }
1108
1109         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1110         if (rc)
1111                 return rc;
1112
1113         if (adapter->rx_scrq) {
1114                 for (i = 0; i < adapter->req_rx_queues; i++) {
1115                         int retries = 10;
1116
1117                         while (pending_scrq(adapter, adapter->rx_scrq[i])) {
1118                                 retries--;
1119                                 mdelay(100);
1120
1121                                 if (retries == 0)
1122                                         break;
1123                         }
1124
1125                         if (adapter->rx_scrq[i]->irq) {
1126                                 netdev_dbg(adapter->netdev,
1127                                            "Disabling rx_scrq[%d] irq\n", i);
1128                                 disable_irq(adapter->rx_scrq[i]->irq);
1129                         }
1130                 }
1131         }
1132
1133         clean_tx_pools(adapter);
1134         adapter->state = VNIC_CLOSED;
1135         return rc;
1136 }
1137
1138 static int ibmvnic_close(struct net_device *netdev)
1139 {
1140         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1141         int rc;
1142
1143         mutex_lock(&adapter->reset_lock);
1144         rc = __ibmvnic_close(netdev);
1145         mutex_unlock(&adapter->reset_lock);
1146
1147         return rc;
1148 }
1149
1150 /**
1151  * build_hdr_data - creates L2/L3/L4 header data buffer
1152  * @hdr_field - bitfield determining needed headers
1153  * @skb - socket buffer
1154  * @hdr_len - array of header lengths
1155  * @tot_len - total length of data
1156  *
1157  * Reads hdr_field to determine which headers are needed by firmware.
1158  * Builds a buffer containing these headers.  Saves individual header
1159  * lengths and total buffer length to be used to build descriptors.
1160  */
1161 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1162                           int *hdr_len, u8 *hdr_data)
1163 {
1164         int len = 0;
1165         u8 *hdr;
1166
1167         hdr_len[0] = sizeof(struct ethhdr);
1168
1169         if (skb->protocol == htons(ETH_P_IP)) {
1170                 hdr_len[1] = ip_hdr(skb)->ihl * 4;
1171                 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1172                         hdr_len[2] = tcp_hdrlen(skb);
1173                 else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1174                         hdr_len[2] = sizeof(struct udphdr);
1175         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1176                 hdr_len[1] = sizeof(struct ipv6hdr);
1177                 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1178                         hdr_len[2] = tcp_hdrlen(skb);
1179                 else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1180                         hdr_len[2] = sizeof(struct udphdr);
1181         } else if (skb->protocol == htons(ETH_P_ARP)) {
1182                 hdr_len[1] = arp_hdr_len(skb->dev);
1183                 hdr_len[2] = 0;
1184         }
1185
1186         memset(hdr_data, 0, 120);
1187         if ((hdr_field >> 6) & 1) {
1188                 hdr = skb_mac_header(skb);
1189                 memcpy(hdr_data, hdr, hdr_len[0]);
1190                 len += hdr_len[0];
1191         }
1192
1193         if ((hdr_field >> 5) & 1) {
1194                 hdr = skb_network_header(skb);
1195                 memcpy(hdr_data + len, hdr, hdr_len[1]);
1196                 len += hdr_len[1];
1197         }
1198
1199         if ((hdr_field >> 4) & 1) {
1200                 hdr = skb_transport_header(skb);
1201                 memcpy(hdr_data + len, hdr, hdr_len[2]);
1202                 len += hdr_len[2];
1203         }
1204         return len;
1205 }
1206
1207 /**
1208  * create_hdr_descs - create header and header extension descriptors
1209  * @hdr_field - bitfield determining needed headers
1210  * @data - buffer containing header data
1211  * @len - length of data buffer
1212  * @hdr_len - array of individual header lengths
1213  * @scrq_arr - descriptor array
1214  *
1215  * Creates header and, if needed, header extension descriptors and
1216  * places them in a descriptor array, scrq_arr
1217  */
1218
1219 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1220                             union sub_crq *scrq_arr)
1221 {
1222         union sub_crq hdr_desc;
1223         int tmp_len = len;
1224         int num_descs = 0;
1225         u8 *data, *cur;
1226         int tmp;
1227
1228         while (tmp_len > 0) {
1229                 cur = hdr_data + len - tmp_len;
1230
1231                 memset(&hdr_desc, 0, sizeof(hdr_desc));
1232                 if (cur != hdr_data) {
1233                         data = hdr_desc.hdr_ext.data;
1234                         tmp = tmp_len > 29 ? 29 : tmp_len;
1235                         hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1236                         hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1237                         hdr_desc.hdr_ext.len = tmp;
1238                 } else {
1239                         data = hdr_desc.hdr.data;
1240                         tmp = tmp_len > 24 ? 24 : tmp_len;
1241                         hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1242                         hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1243                         hdr_desc.hdr.len = tmp;
1244                         hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1245                         hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1246                         hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1247                         hdr_desc.hdr.flag = hdr_field << 1;
1248                 }
1249                 memcpy(data, cur, tmp);
1250                 tmp_len -= tmp;
1251                 *scrq_arr = hdr_desc;
1252                 scrq_arr++;
1253                 num_descs++;
1254         }
1255
1256         return num_descs;
1257 }
1258
1259 /**
1260  * build_hdr_descs_arr - build a header descriptor array
1261  * @skb - socket buffer
1262  * @num_entries - number of descriptors to be sent
1263  * @subcrq - first TX descriptor
1264  * @hdr_field - bit field determining which headers will be sent
1265  *
1266  * This function will build a TX descriptor array with applicable
1267  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1268  */
1269
1270 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1271                                 int *num_entries, u8 hdr_field)
1272 {
1273         int hdr_len[3] = {0, 0, 0};
1274         int tot_len;
1275         u8 *hdr_data = txbuff->hdr_data;
1276
1277         tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1278                                  txbuff->hdr_data);
1279         *num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1280                          txbuff->indir_arr + 1);
1281 }
1282
1283 static int ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1284 {
1285         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1286         int queue_num = skb_get_queue_mapping(skb);
1287         u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1288         struct device *dev = &adapter->vdev->dev;
1289         struct ibmvnic_tx_buff *tx_buff = NULL;
1290         struct ibmvnic_sub_crq_queue *tx_scrq;
1291         struct ibmvnic_tx_pool *tx_pool;
1292         unsigned int tx_send_failed = 0;
1293         unsigned int tx_map_failed = 0;
1294         unsigned int tx_dropped = 0;
1295         unsigned int tx_packets = 0;
1296         unsigned int tx_bytes = 0;
1297         dma_addr_t data_dma_addr;
1298         struct netdev_queue *txq;
1299         unsigned long lpar_rc;
1300         union sub_crq tx_crq;
1301         unsigned int offset;
1302         int num_entries = 1;
1303         unsigned char *dst;
1304         u64 *handle_array;
1305         int index = 0;
1306         u8 proto = 0;
1307         int ret = 0;
1308
1309         if (adapter->resetting) {
1310                 if (!netif_subqueue_stopped(netdev, skb))
1311                         netif_stop_subqueue(netdev, queue_num);
1312                 dev_kfree_skb_any(skb);
1313
1314                 tx_send_failed++;
1315                 tx_dropped++;
1316                 ret = NETDEV_TX_OK;
1317                 goto out;
1318         }
1319
1320         tx_pool = &adapter->tx_pool[queue_num];
1321         tx_scrq = adapter->tx_scrq[queue_num];
1322         txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1323         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1324                 be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1325
1326         index = tx_pool->free_map[tx_pool->consumer_index];
1327
1328         if (skb_is_gso(skb)) {
1329                 offset = tx_pool->tso_index * IBMVNIC_TSO_BUF_SZ;
1330                 dst = tx_pool->tso_ltb.buff + offset;
1331                 memset(dst, 0, IBMVNIC_TSO_BUF_SZ);
1332                 data_dma_addr = tx_pool->tso_ltb.addr + offset;
1333                 tx_pool->tso_index++;
1334                 if (tx_pool->tso_index == IBMVNIC_TSO_BUFS)
1335                         tx_pool->tso_index = 0;
1336         } else {
1337                 offset = index * adapter->req_mtu;
1338                 dst = tx_pool->long_term_buff.buff + offset;
1339                 memset(dst, 0, adapter->req_mtu);
1340                 data_dma_addr = tx_pool->long_term_buff.addr + offset;
1341         }
1342
1343         if (skb_shinfo(skb)->nr_frags) {
1344                 int cur, i;
1345
1346                 /* Copy the head */
1347                 skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1348                 cur = skb_headlen(skb);
1349
1350                 /* Copy the frags */
1351                 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1352                         const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1353
1354                         memcpy(dst + cur,
1355                                page_address(skb_frag_page(frag)) +
1356                                frag->page_offset, skb_frag_size(frag));
1357                         cur += skb_frag_size(frag);
1358                 }
1359         } else {
1360                 skb_copy_from_linear_data(skb, dst, skb->len);
1361         }
1362
1363         tx_pool->consumer_index =
1364             (tx_pool->consumer_index + 1) %
1365                 adapter->req_tx_entries_per_subcrq;
1366
1367         tx_buff = &tx_pool->tx_buff[index];
1368         tx_buff->skb = skb;
1369         tx_buff->data_dma[0] = data_dma_addr;
1370         tx_buff->data_len[0] = skb->len;
1371         tx_buff->index = index;
1372         tx_buff->pool_index = queue_num;
1373         tx_buff->last_frag = true;
1374
1375         memset(&tx_crq, 0, sizeof(tx_crq));
1376         tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1377         tx_crq.v1.type = IBMVNIC_TX_DESC;
1378         tx_crq.v1.n_crq_elem = 1;
1379         tx_crq.v1.n_sge = 1;
1380         tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1381         tx_crq.v1.correlator = cpu_to_be32(index);
1382         if (skb_is_gso(skb))
1383                 tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->tso_ltb.map_id);
1384         else
1385                 tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1386         tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1387         tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1388
1389         if (adapter->vlan_header_insertion) {
1390                 tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1391                 tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1392         }
1393
1394         if (skb->protocol == htons(ETH_P_IP)) {
1395                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1396                 proto = ip_hdr(skb)->protocol;
1397         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1398                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1399                 proto = ipv6_hdr(skb)->nexthdr;
1400         }
1401
1402         if (proto == IPPROTO_TCP)
1403                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1404         else if (proto == IPPROTO_UDP)
1405                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1406
1407         if (skb->ip_summed == CHECKSUM_PARTIAL) {
1408                 tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1409                 hdrs += 2;
1410         }
1411         if (skb_is_gso(skb)) {
1412                 tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1413                 tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1414                 hdrs += 2;
1415         }
1416         /* determine if l2/3/4 headers are sent to firmware */
1417         if ((*hdrs >> 7) & 1 &&
1418             (skb->protocol == htons(ETH_P_IP) ||
1419              skb->protocol == htons(ETH_P_IPV6) ||
1420              skb->protocol == htons(ETH_P_ARP))) {
1421                 build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1422                 tx_crq.v1.n_crq_elem = num_entries;
1423                 tx_buff->indir_arr[0] = tx_crq;
1424                 tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1425                                                     sizeof(tx_buff->indir_arr),
1426                                                     DMA_TO_DEVICE);
1427                 if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1428                         dev_kfree_skb_any(skb);
1429                         tx_buff->skb = NULL;
1430                         if (!firmware_has_feature(FW_FEATURE_CMO))
1431                                 dev_err(dev, "tx: unable to map descriptor array\n");
1432                         tx_map_failed++;
1433                         tx_dropped++;
1434                         ret = NETDEV_TX_OK;
1435                         goto out;
1436                 }
1437                 lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1438                                                (u64)tx_buff->indir_dma,
1439                                                (u64)num_entries);
1440         } else {
1441                 lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1442                                       &tx_crq);
1443         }
1444         if (lpar_rc != H_SUCCESS) {
1445                 dev_err(dev, "tx failed with code %ld\n", lpar_rc);
1446
1447                 if (tx_pool->consumer_index == 0)
1448                         tx_pool->consumer_index =
1449                                 adapter->req_tx_entries_per_subcrq - 1;
1450                 else
1451                         tx_pool->consumer_index--;
1452
1453                 dev_kfree_skb_any(skb);
1454                 tx_buff->skb = NULL;
1455
1456                 if (lpar_rc == H_CLOSED) {
1457                         /* Disable TX and report carrier off if queue is closed.
1458                          * Firmware guarantees that a signal will be sent to the
1459                          * driver, triggering a reset or some other action.
1460                          */
1461                         netif_tx_stop_all_queues(netdev);
1462                         netif_carrier_off(netdev);
1463                 }
1464
1465                 tx_send_failed++;
1466                 tx_dropped++;
1467                 ret = NETDEV_TX_OK;
1468                 goto out;
1469         }
1470
1471         if (atomic_inc_return(&tx_scrq->used)
1472                                         >= adapter->req_tx_entries_per_subcrq) {
1473                 netdev_info(netdev, "Stopping queue %d\n", queue_num);
1474                 netif_stop_subqueue(netdev, queue_num);
1475         }
1476
1477         tx_packets++;
1478         tx_bytes += skb->len;
1479         txq->trans_start = jiffies;
1480         ret = NETDEV_TX_OK;
1481
1482 out:
1483         netdev->stats.tx_dropped += tx_dropped;
1484         netdev->stats.tx_bytes += tx_bytes;
1485         netdev->stats.tx_packets += tx_packets;
1486         adapter->tx_send_failed += tx_send_failed;
1487         adapter->tx_map_failed += tx_map_failed;
1488         adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1489         adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1490         adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1491
1492         return ret;
1493 }
1494
1495 static void ibmvnic_set_multi(struct net_device *netdev)
1496 {
1497         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1498         struct netdev_hw_addr *ha;
1499         union ibmvnic_crq crq;
1500
1501         memset(&crq, 0, sizeof(crq));
1502         crq.request_capability.first = IBMVNIC_CRQ_CMD;
1503         crq.request_capability.cmd = REQUEST_CAPABILITY;
1504
1505         if (netdev->flags & IFF_PROMISC) {
1506                 if (!adapter->promisc_supported)
1507                         return;
1508         } else {
1509                 if (netdev->flags & IFF_ALLMULTI) {
1510                         /* Accept all multicast */
1511                         memset(&crq, 0, sizeof(crq));
1512                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1513                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1514                         crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1515                         ibmvnic_send_crq(adapter, &crq);
1516                 } else if (netdev_mc_empty(netdev)) {
1517                         /* Reject all multicast */
1518                         memset(&crq, 0, sizeof(crq));
1519                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1520                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1521                         crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1522                         ibmvnic_send_crq(adapter, &crq);
1523                 } else {
1524                         /* Accept one or more multicast(s) */
1525                         netdev_for_each_mc_addr(ha, netdev) {
1526                                 memset(&crq, 0, sizeof(crq));
1527                                 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1528                                 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1529                                 crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1530                                 ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1531                                                 ha->addr);
1532                                 ibmvnic_send_crq(adapter, &crq);
1533                         }
1534                 }
1535         }
1536 }
1537
1538 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p)
1539 {
1540         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1541         struct sockaddr *addr = p;
1542         union ibmvnic_crq crq;
1543
1544         if (!is_valid_ether_addr(addr->sa_data))
1545                 return -EADDRNOTAVAIL;
1546
1547         memset(&crq, 0, sizeof(crq));
1548         crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1549         crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1550         ether_addr_copy(&crq.change_mac_addr.mac_addr[0], addr->sa_data);
1551         ibmvnic_send_crq(adapter, &crq);
1552         /* netdev->dev_addr is changed in handle_change_mac_rsp function */
1553         return 0;
1554 }
1555
1556 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1557 {
1558         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1559         struct sockaddr *addr = p;
1560
1561         if (adapter->state == VNIC_PROBED) {
1562                 memcpy(&adapter->desired.mac, addr, sizeof(struct sockaddr));
1563                 adapter->mac_change_pending = true;
1564                 return 0;
1565         }
1566
1567         __ibmvnic_set_mac(netdev, addr);
1568
1569         return 0;
1570 }
1571
1572 /**
1573  * do_reset returns zero if we are able to keep processing reset events, or
1574  * non-zero if we hit a fatal error and must halt.
1575  */
1576 static int do_reset(struct ibmvnic_adapter *adapter,
1577                     struct ibmvnic_rwi *rwi, u32 reset_state)
1578 {
1579         u64 old_num_rx_queues, old_num_tx_queues;
1580         struct net_device *netdev = adapter->netdev;
1581         int i, rc;
1582
1583         netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1584                    rwi->reset_reason);
1585
1586         netif_carrier_off(netdev);
1587         adapter->reset_reason = rwi->reset_reason;
1588
1589         old_num_rx_queues = adapter->req_rx_queues;
1590         old_num_tx_queues = adapter->req_tx_queues;
1591
1592         if (rwi->reset_reason == VNIC_RESET_MOBILITY) {
1593                 rc = ibmvnic_reenable_crq_queue(adapter);
1594                 if (rc)
1595                         return 0;
1596         }
1597
1598         rc = __ibmvnic_close(netdev);
1599         if (rc)
1600                 return rc;
1601
1602         if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1603             adapter->wait_for_reset) {
1604                 release_resources(adapter);
1605                 release_sub_crqs(adapter);
1606                 release_crq_queue(adapter);
1607         }
1608
1609         if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1610                 /* remove the closed state so when we call open it appears
1611                  * we are coming from the probed state.
1612                  */
1613                 adapter->state = VNIC_PROBED;
1614
1615                 rc = ibmvnic_init(adapter);
1616                 if (rc)
1617                         return IBMVNIC_INIT_FAILED;
1618
1619                 /* If the adapter was in PROBE state prior to the reset,
1620                  * exit here.
1621                  */
1622                 if (reset_state == VNIC_PROBED)
1623                         return 0;
1624
1625                 rc = ibmvnic_login(netdev);
1626                 if (rc) {
1627                         adapter->state = VNIC_PROBED;
1628                         return 0;
1629                 }
1630
1631                 if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1632                     adapter->wait_for_reset) {
1633                         rc = init_resources(adapter);
1634                         if (rc)
1635                                 return rc;
1636                 } else if (adapter->req_rx_queues != old_num_rx_queues ||
1637                            adapter->req_tx_queues != old_num_tx_queues) {
1638                         release_rx_pools(adapter);
1639                         release_tx_pools(adapter);
1640                         init_rx_pools(netdev);
1641                         init_tx_pools(netdev);
1642                 } else {
1643                         rc = reset_tx_pools(adapter);
1644                         if (rc)
1645                                 return rc;
1646
1647                         rc = reset_rx_pools(adapter);
1648                         if (rc)
1649                                 return rc;
1650
1651                         if (reset_state == VNIC_CLOSED)
1652                                 return 0;
1653                 }
1654         }
1655
1656         rc = __ibmvnic_open(netdev);
1657         if (rc) {
1658                 if (list_empty(&adapter->rwi_list))
1659                         adapter->state = VNIC_CLOSED;
1660                 else
1661                         adapter->state = reset_state;
1662
1663                 return 0;
1664         }
1665
1666         netif_carrier_on(netdev);
1667
1668         /* kick napi */
1669         for (i = 0; i < adapter->req_rx_queues; i++)
1670                 napi_schedule(&adapter->napi[i]);
1671
1672         if (adapter->reset_reason != VNIC_RESET_FAILOVER)
1673                 netdev_notify_peers(netdev);
1674
1675         return 0;
1676 }
1677
1678 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1679 {
1680         struct ibmvnic_rwi *rwi;
1681
1682         mutex_lock(&adapter->rwi_lock);
1683
1684         if (!list_empty(&adapter->rwi_list)) {
1685                 rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1686                                        list);
1687                 list_del(&rwi->list);
1688         } else {
1689                 rwi = NULL;
1690         }
1691
1692         mutex_unlock(&adapter->rwi_lock);
1693         return rwi;
1694 }
1695
1696 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1697 {
1698         struct ibmvnic_rwi *rwi;
1699
1700         rwi = get_next_rwi(adapter);
1701         while (rwi) {
1702                 kfree(rwi);
1703                 rwi = get_next_rwi(adapter);
1704         }
1705 }
1706
1707 static void __ibmvnic_reset(struct work_struct *work)
1708 {
1709         struct ibmvnic_rwi *rwi;
1710         struct ibmvnic_adapter *adapter;
1711         struct net_device *netdev;
1712         u32 reset_state;
1713         int rc = 0;
1714
1715         adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1716         netdev = adapter->netdev;
1717
1718         mutex_lock(&adapter->reset_lock);
1719         adapter->resetting = true;
1720         reset_state = adapter->state;
1721
1722         rwi = get_next_rwi(adapter);
1723         while (rwi) {
1724                 rc = do_reset(adapter, rwi, reset_state);
1725                 kfree(rwi);
1726                 if (rc && rc != IBMVNIC_INIT_FAILED)
1727                         break;
1728
1729                 rwi = get_next_rwi(adapter);
1730         }
1731
1732         if (adapter->wait_for_reset) {
1733                 adapter->wait_for_reset = false;
1734                 adapter->reset_done_rc = rc;
1735                 complete(&adapter->reset_done);
1736         }
1737
1738         if (rc) {
1739                 netdev_dbg(adapter->netdev, "Reset failed\n");
1740                 free_all_rwi(adapter);
1741                 mutex_unlock(&adapter->reset_lock);
1742                 return;
1743         }
1744
1745         adapter->resetting = false;
1746         mutex_unlock(&adapter->reset_lock);
1747 }
1748
1749 static void ibmvnic_reset(struct ibmvnic_adapter *adapter,
1750                           enum ibmvnic_reset_reason reason)
1751 {
1752         struct ibmvnic_rwi *rwi, *tmp;
1753         struct net_device *netdev = adapter->netdev;
1754         struct list_head *entry;
1755
1756         if (adapter->state == VNIC_REMOVING ||
1757             adapter->state == VNIC_REMOVED) {
1758                 netdev_dbg(netdev, "Adapter removing, skipping reset\n");
1759                 return;
1760         }
1761
1762         if (adapter->state == VNIC_PROBING) {
1763                 netdev_warn(netdev, "Adapter reset during probe\n");
1764                 adapter->init_done_rc = EAGAIN;
1765                 return;
1766         }
1767
1768         mutex_lock(&adapter->rwi_lock);
1769
1770         list_for_each(entry, &adapter->rwi_list) {
1771                 tmp = list_entry(entry, struct ibmvnic_rwi, list);
1772                 if (tmp->reset_reason == reason) {
1773                         netdev_dbg(netdev, "Skipping matching reset\n");
1774                         mutex_unlock(&adapter->rwi_lock);
1775                         return;
1776                 }
1777         }
1778
1779         rwi = kzalloc(sizeof(*rwi), GFP_KERNEL);
1780         if (!rwi) {
1781                 mutex_unlock(&adapter->rwi_lock);
1782                 ibmvnic_close(netdev);
1783                 return;
1784         }
1785
1786         rwi->reset_reason = reason;
1787         list_add_tail(&rwi->list, &adapter->rwi_list);
1788         mutex_unlock(&adapter->rwi_lock);
1789
1790         netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
1791         schedule_work(&adapter->ibmvnic_reset);
1792 }
1793
1794 static void ibmvnic_tx_timeout(struct net_device *dev)
1795 {
1796         struct ibmvnic_adapter *adapter = netdev_priv(dev);
1797
1798         ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
1799 }
1800
1801 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
1802                                   struct ibmvnic_rx_buff *rx_buff)
1803 {
1804         struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
1805
1806         rx_buff->skb = NULL;
1807
1808         pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
1809         pool->next_alloc = (pool->next_alloc + 1) % pool->size;
1810
1811         atomic_dec(&pool->available);
1812 }
1813
1814 static int ibmvnic_poll(struct napi_struct *napi, int budget)
1815 {
1816         struct net_device *netdev = napi->dev;
1817         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1818         int scrq_num = (int)(napi - adapter->napi);
1819         int frames_processed = 0;
1820
1821 restart_poll:
1822         while (frames_processed < budget) {
1823                 struct sk_buff *skb;
1824                 struct ibmvnic_rx_buff *rx_buff;
1825                 union sub_crq *next;
1826                 u32 length;
1827                 u16 offset;
1828                 u8 flags = 0;
1829
1830                 if (unlikely(adapter->resetting)) {
1831                         enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1832                         napi_complete_done(napi, frames_processed);
1833                         return frames_processed;
1834                 }
1835
1836                 if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
1837                         break;
1838                 next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
1839                 rx_buff =
1840                     (struct ibmvnic_rx_buff *)be64_to_cpu(next->
1841                                                           rx_comp.correlator);
1842                 /* do error checking */
1843                 if (next->rx_comp.rc) {
1844                         netdev_dbg(netdev, "rx buffer returned with rc %x\n",
1845                                    be16_to_cpu(next->rx_comp.rc));
1846                         /* free the entry */
1847                         next->rx_comp.first = 0;
1848                         remove_buff_from_pool(adapter, rx_buff);
1849                         continue;
1850                 }
1851
1852                 length = be32_to_cpu(next->rx_comp.len);
1853                 offset = be16_to_cpu(next->rx_comp.off_frame_data);
1854                 flags = next->rx_comp.flags;
1855                 skb = rx_buff->skb;
1856                 skb_copy_to_linear_data(skb, rx_buff->data + offset,
1857                                         length);
1858
1859                 /* VLAN Header has been stripped by the system firmware and
1860                  * needs to be inserted by the driver
1861                  */
1862                 if (adapter->rx_vlan_header_insertion &&
1863                     (flags & IBMVNIC_VLAN_STRIPPED))
1864                         __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
1865                                                ntohs(next->rx_comp.vlan_tci));
1866
1867                 /* free the entry */
1868                 next->rx_comp.first = 0;
1869                 remove_buff_from_pool(adapter, rx_buff);
1870
1871                 skb_put(skb, length);
1872                 skb->protocol = eth_type_trans(skb, netdev);
1873                 skb_record_rx_queue(skb, scrq_num);
1874
1875                 if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
1876                     flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
1877                         skb->ip_summed = CHECKSUM_UNNECESSARY;
1878                 }
1879
1880                 length = skb->len;
1881                 napi_gro_receive(napi, skb); /* send it up */
1882                 netdev->stats.rx_packets++;
1883                 netdev->stats.rx_bytes += length;
1884                 adapter->rx_stats_buffers[scrq_num].packets++;
1885                 adapter->rx_stats_buffers[scrq_num].bytes += length;
1886                 frames_processed++;
1887         }
1888
1889         if (adapter->state != VNIC_CLOSING)
1890                 replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
1891
1892         if (frames_processed < budget) {
1893                 enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1894                 napi_complete_done(napi, frames_processed);
1895                 if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
1896                     napi_reschedule(napi)) {
1897                         disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1898                         goto restart_poll;
1899                 }
1900         }
1901         return frames_processed;
1902 }
1903
1904 #ifdef CONFIG_NET_POLL_CONTROLLER
1905 static void ibmvnic_netpoll_controller(struct net_device *dev)
1906 {
1907         struct ibmvnic_adapter *adapter = netdev_priv(dev);
1908         int i;
1909
1910         replenish_pools(netdev_priv(dev));
1911         for (i = 0; i < adapter->req_rx_queues; i++)
1912                 ibmvnic_interrupt_rx(adapter->rx_scrq[i]->irq,
1913                                      adapter->rx_scrq[i]);
1914 }
1915 #endif
1916
1917 static int wait_for_reset(struct ibmvnic_adapter *adapter)
1918 {
1919         adapter->fallback.mtu = adapter->req_mtu;
1920         adapter->fallback.rx_queues = adapter->req_rx_queues;
1921         adapter->fallback.tx_queues = adapter->req_tx_queues;
1922         adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
1923         adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
1924
1925         init_completion(&adapter->reset_done);
1926         ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
1927         adapter->wait_for_reset = true;
1928         wait_for_completion(&adapter->reset_done);
1929
1930         if (adapter->reset_done_rc) {
1931                 adapter->desired.mtu = adapter->fallback.mtu;
1932                 adapter->desired.rx_queues = adapter->fallback.rx_queues;
1933                 adapter->desired.tx_queues = adapter->fallback.tx_queues;
1934                 adapter->desired.rx_entries = adapter->fallback.rx_entries;
1935                 adapter->desired.tx_entries = adapter->fallback.tx_entries;
1936
1937                 init_completion(&adapter->reset_done);
1938                 ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
1939                 wait_for_completion(&adapter->reset_done);
1940         }
1941         adapter->wait_for_reset = false;
1942
1943         return adapter->reset_done_rc;
1944 }
1945
1946 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
1947 {
1948         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1949
1950         adapter->desired.mtu = new_mtu + ETH_HLEN;
1951
1952         return wait_for_reset(adapter);
1953 }
1954
1955 static const struct net_device_ops ibmvnic_netdev_ops = {
1956         .ndo_open               = ibmvnic_open,
1957         .ndo_stop               = ibmvnic_close,
1958         .ndo_start_xmit         = ibmvnic_xmit,
1959         .ndo_set_rx_mode        = ibmvnic_set_multi,
1960         .ndo_set_mac_address    = ibmvnic_set_mac,
1961         .ndo_validate_addr      = eth_validate_addr,
1962         .ndo_tx_timeout         = ibmvnic_tx_timeout,
1963 #ifdef CONFIG_NET_POLL_CONTROLLER
1964         .ndo_poll_controller    = ibmvnic_netpoll_controller,
1965 #endif
1966         .ndo_change_mtu         = ibmvnic_change_mtu,
1967 };
1968
1969 /* ethtool functions */
1970
1971 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
1972                                       struct ethtool_link_ksettings *cmd)
1973 {
1974         u32 supported, advertising;
1975
1976         supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg |
1977                           SUPPORTED_FIBRE);
1978         advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg |
1979                             ADVERTISED_FIBRE);
1980         cmd->base.speed = SPEED_1000;
1981         cmd->base.duplex = DUPLEX_FULL;
1982         cmd->base.port = PORT_FIBRE;
1983         cmd->base.phy_address = 0;
1984         cmd->base.autoneg = AUTONEG_ENABLE;
1985
1986         ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
1987                                                 supported);
1988         ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.advertising,
1989                                                 advertising);
1990
1991         return 0;
1992 }
1993
1994 static void ibmvnic_get_drvinfo(struct net_device *netdev,
1995                                 struct ethtool_drvinfo *info)
1996 {
1997         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1998
1999         strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2000         strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2001         strlcpy(info->fw_version, adapter->fw_version,
2002                 sizeof(info->fw_version));
2003 }
2004
2005 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2006 {
2007         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2008
2009         return adapter->msg_enable;
2010 }
2011
2012 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2013 {
2014         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2015
2016         adapter->msg_enable = data;
2017 }
2018
2019 static u32 ibmvnic_get_link(struct net_device *netdev)
2020 {
2021         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2022
2023         /* Don't need to send a query because we request a logical link up at
2024          * init and then we wait for link state indications
2025          */
2026         return adapter->logical_link_state;
2027 }
2028
2029 static void ibmvnic_get_ringparam(struct net_device *netdev,
2030                                   struct ethtool_ringparam *ring)
2031 {
2032         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2033
2034         ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2035         ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2036         ring->rx_mini_max_pending = 0;
2037         ring->rx_jumbo_max_pending = 0;
2038         ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2039         ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2040         ring->rx_mini_pending = 0;
2041         ring->rx_jumbo_pending = 0;
2042 }
2043
2044 static int ibmvnic_set_ringparam(struct net_device *netdev,
2045                                  struct ethtool_ringparam *ring)
2046 {
2047         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2048
2049         if (ring->rx_pending > adapter->max_rx_add_entries_per_subcrq  ||
2050             ring->tx_pending > adapter->max_tx_entries_per_subcrq) {
2051                 netdev_err(netdev, "Invalid request.\n");
2052                 netdev_err(netdev, "Max tx buffers = %llu\n",
2053                            adapter->max_rx_add_entries_per_subcrq);
2054                 netdev_err(netdev, "Max rx buffers = %llu\n",
2055                            adapter->max_tx_entries_per_subcrq);
2056                 return -EINVAL;
2057         }
2058
2059         adapter->desired.rx_entries = ring->rx_pending;
2060         adapter->desired.tx_entries = ring->tx_pending;
2061
2062         return wait_for_reset(adapter);
2063 }
2064
2065 static void ibmvnic_get_channels(struct net_device *netdev,
2066                                  struct ethtool_channels *channels)
2067 {
2068         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2069
2070         channels->max_rx = adapter->max_rx_queues;
2071         channels->max_tx = adapter->max_tx_queues;
2072         channels->max_other = 0;
2073         channels->max_combined = 0;
2074         channels->rx_count = adapter->req_rx_queues;
2075         channels->tx_count = adapter->req_tx_queues;
2076         channels->other_count = 0;
2077         channels->combined_count = 0;
2078 }
2079
2080 static int ibmvnic_set_channels(struct net_device *netdev,
2081                                 struct ethtool_channels *channels)
2082 {
2083         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2084
2085         adapter->desired.rx_queues = channels->rx_count;
2086         adapter->desired.tx_queues = channels->tx_count;
2087
2088         return wait_for_reset(adapter);
2089 }
2090
2091 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2092 {
2093         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2094         int i;
2095
2096         if (stringset != ETH_SS_STATS)
2097                 return;
2098
2099         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++, data += ETH_GSTRING_LEN)
2100                 memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2101
2102         for (i = 0; i < adapter->req_tx_queues; i++) {
2103                 snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2104                 data += ETH_GSTRING_LEN;
2105
2106                 snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2107                 data += ETH_GSTRING_LEN;
2108
2109                 snprintf(data, ETH_GSTRING_LEN, "tx%d_dropped_packets", i);
2110                 data += ETH_GSTRING_LEN;
2111         }
2112
2113         for (i = 0; i < adapter->req_rx_queues; i++) {
2114                 snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2115                 data += ETH_GSTRING_LEN;
2116
2117                 snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2118                 data += ETH_GSTRING_LEN;
2119
2120                 snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2121                 data += ETH_GSTRING_LEN;
2122         }
2123 }
2124
2125 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2126 {
2127         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2128
2129         switch (sset) {
2130         case ETH_SS_STATS:
2131                 return ARRAY_SIZE(ibmvnic_stats) +
2132                        adapter->req_tx_queues * NUM_TX_STATS +
2133                        adapter->req_rx_queues * NUM_RX_STATS;
2134         default:
2135                 return -EOPNOTSUPP;
2136         }
2137 }
2138
2139 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2140                                       struct ethtool_stats *stats, u64 *data)
2141 {
2142         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2143         union ibmvnic_crq crq;
2144         int i, j;
2145
2146         memset(&crq, 0, sizeof(crq));
2147         crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2148         crq.request_statistics.cmd = REQUEST_STATISTICS;
2149         crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2150         crq.request_statistics.len =
2151             cpu_to_be32(sizeof(struct ibmvnic_statistics));
2152
2153         /* Wait for data to be written */
2154         init_completion(&adapter->stats_done);
2155         ibmvnic_send_crq(adapter, &crq);
2156         wait_for_completion(&adapter->stats_done);
2157
2158         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2159                 data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2160                                                 ibmvnic_stats[i].offset));
2161
2162         for (j = 0; j < adapter->req_tx_queues; j++) {
2163                 data[i] = adapter->tx_stats_buffers[j].packets;
2164                 i++;
2165                 data[i] = adapter->tx_stats_buffers[j].bytes;
2166                 i++;
2167                 data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2168                 i++;
2169         }
2170
2171         for (j = 0; j < adapter->req_rx_queues; j++) {
2172                 data[i] = adapter->rx_stats_buffers[j].packets;
2173                 i++;
2174                 data[i] = adapter->rx_stats_buffers[j].bytes;
2175                 i++;
2176                 data[i] = adapter->rx_stats_buffers[j].interrupts;
2177                 i++;
2178         }
2179 }
2180
2181 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2182         .get_drvinfo            = ibmvnic_get_drvinfo,
2183         .get_msglevel           = ibmvnic_get_msglevel,
2184         .set_msglevel           = ibmvnic_set_msglevel,
2185         .get_link               = ibmvnic_get_link,
2186         .get_ringparam          = ibmvnic_get_ringparam,
2187         .set_ringparam          = ibmvnic_set_ringparam,
2188         .get_channels           = ibmvnic_get_channels,
2189         .set_channels           = ibmvnic_set_channels,
2190         .get_strings            = ibmvnic_get_strings,
2191         .get_sset_count         = ibmvnic_get_sset_count,
2192         .get_ethtool_stats      = ibmvnic_get_ethtool_stats,
2193         .get_link_ksettings     = ibmvnic_get_link_ksettings,
2194 };
2195
2196 /* Routines for managing CRQs/sCRQs  */
2197
2198 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2199                                    struct ibmvnic_sub_crq_queue *scrq)
2200 {
2201         int rc;
2202
2203         if (scrq->irq) {
2204                 free_irq(scrq->irq, scrq);
2205                 irq_dispose_mapping(scrq->irq);
2206                 scrq->irq = 0;
2207         }
2208
2209         memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2210         scrq->cur = 0;
2211
2212         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2213                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2214         return rc;
2215 }
2216
2217 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2218 {
2219         int i, rc;
2220
2221         for (i = 0; i < adapter->req_tx_queues; i++) {
2222                 netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2223                 rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2224                 if (rc)
2225                         return rc;
2226         }
2227
2228         for (i = 0; i < adapter->req_rx_queues; i++) {
2229                 netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2230                 rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2231                 if (rc)
2232                         return rc;
2233         }
2234
2235         return rc;
2236 }
2237
2238 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2239                                   struct ibmvnic_sub_crq_queue *scrq)
2240 {
2241         struct device *dev = &adapter->vdev->dev;
2242         long rc;
2243
2244         netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2245
2246         /* Close the sub-crqs */
2247         do {
2248                 rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2249                                         adapter->vdev->unit_address,
2250                                         scrq->crq_num);
2251         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2252
2253         if (rc) {
2254                 netdev_err(adapter->netdev,
2255                            "Failed to release sub-CRQ %16lx, rc = %ld\n",
2256                            scrq->crq_num, rc);
2257         }
2258
2259         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2260                          DMA_BIDIRECTIONAL);
2261         free_pages((unsigned long)scrq->msgs, 2);
2262         kfree(scrq);
2263 }
2264
2265 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2266                                                         *adapter)
2267 {
2268         struct device *dev = &adapter->vdev->dev;
2269         struct ibmvnic_sub_crq_queue *scrq;
2270         int rc;
2271
2272         scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2273         if (!scrq)
2274                 return NULL;
2275
2276         scrq->msgs =
2277                 (union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2278         if (!scrq->msgs) {
2279                 dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2280                 goto zero_page_failed;
2281         }
2282
2283         scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2284                                          DMA_BIDIRECTIONAL);
2285         if (dma_mapping_error(dev, scrq->msg_token)) {
2286                 dev_warn(dev, "Couldn't map crq queue messages page\n");
2287                 goto map_failed;
2288         }
2289
2290         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2291                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2292
2293         if (rc == H_RESOURCE)
2294                 rc = ibmvnic_reset_crq(adapter);
2295
2296         if (rc == H_CLOSED) {
2297                 dev_warn(dev, "Partner adapter not ready, waiting.\n");
2298         } else if (rc) {
2299                 dev_warn(dev, "Error %d registering sub-crq\n", rc);
2300                 goto reg_failed;
2301         }
2302
2303         scrq->adapter = adapter;
2304         scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2305         spin_lock_init(&scrq->lock);
2306
2307         netdev_dbg(adapter->netdev,
2308                    "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2309                    scrq->crq_num, scrq->hw_irq, scrq->irq);
2310
2311         return scrq;
2312
2313 reg_failed:
2314         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2315                          DMA_BIDIRECTIONAL);
2316 map_failed:
2317         free_pages((unsigned long)scrq->msgs, 2);
2318 zero_page_failed:
2319         kfree(scrq);
2320
2321         return NULL;
2322 }
2323
2324 static void release_sub_crqs(struct ibmvnic_adapter *adapter)
2325 {
2326         int i;
2327
2328         if (adapter->tx_scrq) {
2329                 for (i = 0; i < adapter->req_tx_queues; i++) {
2330                         if (!adapter->tx_scrq[i])
2331                                 continue;
2332
2333                         netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2334                                    i);
2335                         if (adapter->tx_scrq[i]->irq) {
2336                                 free_irq(adapter->tx_scrq[i]->irq,
2337                                          adapter->tx_scrq[i]);
2338                                 irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2339                                 adapter->tx_scrq[i]->irq = 0;
2340                         }
2341
2342                         release_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2343                 }
2344
2345                 kfree(adapter->tx_scrq);
2346                 adapter->tx_scrq = NULL;
2347         }
2348
2349         if (adapter->rx_scrq) {
2350                 for (i = 0; i < adapter->req_rx_queues; i++) {
2351                         if (!adapter->rx_scrq[i])
2352                                 continue;
2353
2354                         netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2355                                    i);
2356                         if (adapter->rx_scrq[i]->irq) {
2357                                 free_irq(adapter->rx_scrq[i]->irq,
2358                                          adapter->rx_scrq[i]);
2359                                 irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2360                                 adapter->rx_scrq[i]->irq = 0;
2361                         }
2362
2363                         release_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2364                 }
2365
2366                 kfree(adapter->rx_scrq);
2367                 adapter->rx_scrq = NULL;
2368         }
2369 }
2370
2371 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2372                             struct ibmvnic_sub_crq_queue *scrq)
2373 {
2374         struct device *dev = &adapter->vdev->dev;
2375         unsigned long rc;
2376
2377         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2378                                 H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2379         if (rc)
2380                 dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2381                         scrq->hw_irq, rc);
2382         return rc;
2383 }
2384
2385 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2386                            struct ibmvnic_sub_crq_queue *scrq)
2387 {
2388         struct device *dev = &adapter->vdev->dev;
2389         unsigned long rc;
2390
2391         if (scrq->hw_irq > 0x100000000ULL) {
2392                 dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2393                 return 1;
2394         }
2395
2396         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2397                                 H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2398         if (rc)
2399                 dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
2400                         scrq->hw_irq, rc);
2401         return rc;
2402 }
2403
2404 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
2405                                struct ibmvnic_sub_crq_queue *scrq)
2406 {
2407         struct device *dev = &adapter->vdev->dev;
2408         struct ibmvnic_tx_buff *txbuff;
2409         union sub_crq *next;
2410         int index;
2411         int i, j;
2412         u8 first;
2413
2414 restart_loop:
2415         while (pending_scrq(adapter, scrq)) {
2416                 unsigned int pool = scrq->pool_index;
2417
2418                 next = ibmvnic_next_scrq(adapter, scrq);
2419                 for (i = 0; i < next->tx_comp.num_comps; i++) {
2420                         if (next->tx_comp.rcs[i]) {
2421                                 dev_err(dev, "tx error %x\n",
2422                                         next->tx_comp.rcs[i]);
2423                                 continue;
2424                         }
2425                         index = be32_to_cpu(next->tx_comp.correlators[i]);
2426                         txbuff = &adapter->tx_pool[pool].tx_buff[index];
2427
2428                         for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
2429                                 if (!txbuff->data_dma[j])
2430                                         continue;
2431
2432                                 txbuff->data_dma[j] = 0;
2433                         }
2434                         /* if sub_crq was sent indirectly */
2435                         first = txbuff->indir_arr[0].generic.first;
2436                         if (first == IBMVNIC_CRQ_CMD) {
2437                                 dma_unmap_single(dev, txbuff->indir_dma,
2438                                                  sizeof(txbuff->indir_arr),
2439                                                  DMA_TO_DEVICE);
2440                         }
2441
2442                         if (txbuff->last_frag) {
2443                                 dev_kfree_skb_any(txbuff->skb);
2444                                 txbuff->skb = NULL;
2445                         }
2446
2447                         adapter->tx_pool[pool].free_map[adapter->tx_pool[pool].
2448                                                      producer_index] = index;
2449                         adapter->tx_pool[pool].producer_index =
2450                             (adapter->tx_pool[pool].producer_index + 1) %
2451                             adapter->req_tx_entries_per_subcrq;
2452                 }
2453                 /* remove tx_comp scrq*/
2454                 next->tx_comp.first = 0;
2455
2456                 if (atomic_sub_return(next->tx_comp.num_comps, &scrq->used) <=
2457                     (adapter->req_tx_entries_per_subcrq / 2) &&
2458                     __netif_subqueue_stopped(adapter->netdev,
2459                                              scrq->pool_index)) {
2460                         netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2461                         netdev_info(adapter->netdev, "Started queue %d\n",
2462                                     scrq->pool_index);
2463                 }
2464         }
2465
2466         enable_scrq_irq(adapter, scrq);
2467
2468         if (pending_scrq(adapter, scrq)) {
2469                 disable_scrq_irq(adapter, scrq);
2470                 goto restart_loop;
2471         }
2472
2473         return 0;
2474 }
2475
2476 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2477 {
2478         struct ibmvnic_sub_crq_queue *scrq = instance;
2479         struct ibmvnic_adapter *adapter = scrq->adapter;
2480
2481         disable_scrq_irq(adapter, scrq);
2482         ibmvnic_complete_tx(adapter, scrq);
2483
2484         return IRQ_HANDLED;
2485 }
2486
2487 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2488 {
2489         struct ibmvnic_sub_crq_queue *scrq = instance;
2490         struct ibmvnic_adapter *adapter = scrq->adapter;
2491
2492         /* When booting a kdump kernel we can hit pending interrupts
2493          * prior to completing driver initialization.
2494          */
2495         if (unlikely(adapter->state != VNIC_OPEN))
2496                 return IRQ_NONE;
2497
2498         adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
2499
2500         if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2501                 disable_scrq_irq(adapter, scrq);
2502                 __napi_schedule(&adapter->napi[scrq->scrq_num]);
2503         }
2504
2505         return IRQ_HANDLED;
2506 }
2507
2508 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2509 {
2510         struct device *dev = &adapter->vdev->dev;
2511         struct ibmvnic_sub_crq_queue *scrq;
2512         int i = 0, j = 0;
2513         int rc = 0;
2514
2515         for (i = 0; i < adapter->req_tx_queues; i++) {
2516                 netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
2517                            i);
2518                 scrq = adapter->tx_scrq[i];
2519                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2520
2521                 if (!scrq->irq) {
2522                         rc = -EINVAL;
2523                         dev_err(dev, "Error mapping irq\n");
2524                         goto req_tx_irq_failed;
2525                 }
2526
2527                 rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2528                                  0, "ibmvnic_tx", scrq);
2529
2530                 if (rc) {
2531                         dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2532                                 scrq->irq, rc);
2533                         irq_dispose_mapping(scrq->irq);
2534                         goto req_rx_irq_failed;
2535                 }
2536         }
2537
2538         for (i = 0; i < adapter->req_rx_queues; i++) {
2539                 netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
2540                            i);
2541                 scrq = adapter->rx_scrq[i];
2542                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2543                 if (!scrq->irq) {
2544                         rc = -EINVAL;
2545                         dev_err(dev, "Error mapping irq\n");
2546                         goto req_rx_irq_failed;
2547                 }
2548                 rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2549                                  0, "ibmvnic_rx", scrq);
2550                 if (rc) {
2551                         dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2552                                 scrq->irq, rc);
2553                         irq_dispose_mapping(scrq->irq);
2554                         goto req_rx_irq_failed;
2555                 }
2556         }
2557         return rc;
2558
2559 req_rx_irq_failed:
2560         for (j = 0; j < i; j++) {
2561                 free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2562                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2563         }
2564         i = adapter->req_tx_queues;
2565 req_tx_irq_failed:
2566         for (j = 0; j < i; j++) {
2567                 free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2568                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2569         }
2570         release_sub_crqs(adapter);
2571         return rc;
2572 }
2573
2574 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2575 {
2576         struct device *dev = &adapter->vdev->dev;
2577         struct ibmvnic_sub_crq_queue **allqueues;
2578         int registered_queues = 0;
2579         int total_queues;
2580         int more = 0;
2581         int i;
2582
2583         total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2584
2585         allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2586         if (!allqueues)
2587                 return -1;
2588
2589         for (i = 0; i < total_queues; i++) {
2590                 allqueues[i] = init_sub_crq_queue(adapter);
2591                 if (!allqueues[i]) {
2592                         dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2593                         break;
2594                 }
2595                 registered_queues++;
2596         }
2597
2598         /* Make sure we were able to register the minimum number of queues */
2599         if (registered_queues <
2600             adapter->min_tx_queues + adapter->min_rx_queues) {
2601                 dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
2602                 goto tx_failed;
2603         }
2604
2605         /* Distribute the failed allocated queues*/
2606         for (i = 0; i < total_queues - registered_queues + more ; i++) {
2607                 netdev_dbg(adapter->netdev, "Reducing number of queues\n");
2608                 switch (i % 3) {
2609                 case 0:
2610                         if (adapter->req_rx_queues > adapter->min_rx_queues)
2611                                 adapter->req_rx_queues--;
2612                         else
2613                                 more++;
2614                         break;
2615                 case 1:
2616                         if (adapter->req_tx_queues > adapter->min_tx_queues)
2617                                 adapter->req_tx_queues--;
2618                         else
2619                                 more++;
2620                         break;
2621                 }
2622         }
2623
2624         adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
2625                                    sizeof(*adapter->tx_scrq), GFP_KERNEL);
2626         if (!adapter->tx_scrq)
2627                 goto tx_failed;
2628
2629         for (i = 0; i < adapter->req_tx_queues; i++) {
2630                 adapter->tx_scrq[i] = allqueues[i];
2631                 adapter->tx_scrq[i]->pool_index = i;
2632         }
2633
2634         adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
2635                                    sizeof(*adapter->rx_scrq), GFP_KERNEL);
2636         if (!adapter->rx_scrq)
2637                 goto rx_failed;
2638
2639         for (i = 0; i < adapter->req_rx_queues; i++) {
2640                 adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
2641                 adapter->rx_scrq[i]->scrq_num = i;
2642         }
2643
2644         kfree(allqueues);
2645         return 0;
2646
2647 rx_failed:
2648         kfree(adapter->tx_scrq);
2649         adapter->tx_scrq = NULL;
2650 tx_failed:
2651         for (i = 0; i < registered_queues; i++)
2652                 release_sub_crq_queue(adapter, allqueues[i]);
2653         kfree(allqueues);
2654         return -1;
2655 }
2656
2657 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
2658 {
2659         struct device *dev = &adapter->vdev->dev;
2660         union ibmvnic_crq crq;
2661         int max_entries;
2662
2663         if (!retry) {
2664                 /* Sub-CRQ entries are 32 byte long */
2665                 int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
2666
2667                 if (adapter->min_tx_entries_per_subcrq > entries_page ||
2668                     adapter->min_rx_add_entries_per_subcrq > entries_page) {
2669                         dev_err(dev, "Fatal, invalid entries per sub-crq\n");
2670                         return;
2671                 }
2672
2673                 if (adapter->desired.mtu)
2674                         adapter->req_mtu = adapter->desired.mtu;
2675                 else
2676                         adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
2677
2678                 if (!adapter->desired.tx_entries)
2679                         adapter->desired.tx_entries =
2680                                         adapter->max_tx_entries_per_subcrq;
2681                 if (!adapter->desired.rx_entries)
2682                         adapter->desired.rx_entries =
2683                                         adapter->max_rx_add_entries_per_subcrq;
2684
2685                 max_entries = IBMVNIC_MAX_LTB_SIZE /
2686                               (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
2687
2688                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
2689                         adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
2690                         adapter->desired.tx_entries = max_entries;
2691                 }
2692
2693                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
2694                         adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
2695                         adapter->desired.rx_entries = max_entries;
2696                 }
2697
2698                 if (adapter->desired.tx_entries)
2699                         adapter->req_tx_entries_per_subcrq =
2700                                         adapter->desired.tx_entries;
2701                 else
2702                         adapter->req_tx_entries_per_subcrq =
2703                                         adapter->max_tx_entries_per_subcrq;
2704
2705                 if (adapter->desired.rx_entries)
2706                         adapter->req_rx_add_entries_per_subcrq =
2707                                         adapter->desired.rx_entries;
2708                 else
2709                         adapter->req_rx_add_entries_per_subcrq =
2710                                         adapter->max_rx_add_entries_per_subcrq;
2711
2712                 if (adapter->desired.tx_queues)
2713                         adapter->req_tx_queues =
2714                                         adapter->desired.tx_queues;
2715                 else
2716                         adapter->req_tx_queues =
2717                                         adapter->opt_tx_comp_sub_queues;
2718
2719                 if (adapter->desired.rx_queues)
2720                         adapter->req_rx_queues =
2721                                         adapter->desired.rx_queues;
2722                 else
2723                         adapter->req_rx_queues =
2724                                         adapter->opt_rx_comp_queues;
2725
2726                 adapter->req_rx_add_queues = adapter->max_rx_add_queues;
2727         }
2728
2729         memset(&crq, 0, sizeof(crq));
2730         crq.request_capability.first = IBMVNIC_CRQ_CMD;
2731         crq.request_capability.cmd = REQUEST_CAPABILITY;
2732
2733         crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
2734         crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
2735         atomic_inc(&adapter->running_cap_crqs);
2736         ibmvnic_send_crq(adapter, &crq);
2737
2738         crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
2739         crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
2740         atomic_inc(&adapter->running_cap_crqs);
2741         ibmvnic_send_crq(adapter, &crq);
2742
2743         crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
2744         crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
2745         atomic_inc(&adapter->running_cap_crqs);
2746         ibmvnic_send_crq(adapter, &crq);
2747
2748         crq.request_capability.capability =
2749             cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
2750         crq.request_capability.number =
2751             cpu_to_be64(adapter->req_tx_entries_per_subcrq);
2752         atomic_inc(&adapter->running_cap_crqs);
2753         ibmvnic_send_crq(adapter, &crq);
2754
2755         crq.request_capability.capability =
2756             cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
2757         crq.request_capability.number =
2758             cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
2759         atomic_inc(&adapter->running_cap_crqs);
2760         ibmvnic_send_crq(adapter, &crq);
2761
2762         crq.request_capability.capability = cpu_to_be16(REQ_MTU);
2763         crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
2764         atomic_inc(&adapter->running_cap_crqs);
2765         ibmvnic_send_crq(adapter, &crq);
2766
2767         if (adapter->netdev->flags & IFF_PROMISC) {
2768                 if (adapter->promisc_supported) {
2769                         crq.request_capability.capability =
2770                             cpu_to_be16(PROMISC_REQUESTED);
2771                         crq.request_capability.number = cpu_to_be64(1);
2772                         atomic_inc(&adapter->running_cap_crqs);
2773                         ibmvnic_send_crq(adapter, &crq);
2774                 }
2775         } else {
2776                 crq.request_capability.capability =
2777                     cpu_to_be16(PROMISC_REQUESTED);
2778                 crq.request_capability.number = cpu_to_be64(0);
2779                 atomic_inc(&adapter->running_cap_crqs);
2780                 ibmvnic_send_crq(adapter, &crq);
2781         }
2782 }
2783
2784 static int pending_scrq(struct ibmvnic_adapter *adapter,
2785                         struct ibmvnic_sub_crq_queue *scrq)
2786 {
2787         union sub_crq *entry = &scrq->msgs[scrq->cur];
2788
2789         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
2790                 return 1;
2791         else
2792                 return 0;
2793 }
2794
2795 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
2796                                         struct ibmvnic_sub_crq_queue *scrq)
2797 {
2798         union sub_crq *entry;
2799         unsigned long flags;
2800
2801         spin_lock_irqsave(&scrq->lock, flags);
2802         entry = &scrq->msgs[scrq->cur];
2803         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
2804                 if (++scrq->cur == scrq->size)
2805                         scrq->cur = 0;
2806         } else {
2807                 entry = NULL;
2808         }
2809         spin_unlock_irqrestore(&scrq->lock, flags);
2810
2811         return entry;
2812 }
2813
2814 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
2815 {
2816         struct ibmvnic_crq_queue *queue = &adapter->crq;
2817         union ibmvnic_crq *crq;
2818
2819         crq = &queue->msgs[queue->cur];
2820         if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
2821                 if (++queue->cur == queue->size)
2822                         queue->cur = 0;
2823         } else {
2824                 crq = NULL;
2825         }
2826
2827         return crq;
2828 }
2829
2830 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
2831                        union sub_crq *sub_crq)
2832 {
2833         unsigned int ua = adapter->vdev->unit_address;
2834         struct device *dev = &adapter->vdev->dev;
2835         u64 *u64_crq = (u64 *)sub_crq;
2836         int rc;
2837
2838         netdev_dbg(adapter->netdev,
2839                    "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
2840                    (unsigned long int)cpu_to_be64(remote_handle),
2841                    (unsigned long int)cpu_to_be64(u64_crq[0]),
2842                    (unsigned long int)cpu_to_be64(u64_crq[1]),
2843                    (unsigned long int)cpu_to_be64(u64_crq[2]),
2844                    (unsigned long int)cpu_to_be64(u64_crq[3]));
2845
2846         /* Make sure the hypervisor sees the complete request */
2847         mb();
2848
2849         rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
2850                                 cpu_to_be64(remote_handle),
2851                                 cpu_to_be64(u64_crq[0]),
2852                                 cpu_to_be64(u64_crq[1]),
2853                                 cpu_to_be64(u64_crq[2]),
2854                                 cpu_to_be64(u64_crq[3]));
2855
2856         if (rc) {
2857                 if (rc == H_CLOSED)
2858                         dev_warn(dev, "CRQ Queue closed\n");
2859                 dev_err(dev, "Send error (rc=%d)\n", rc);
2860         }
2861
2862         return rc;
2863 }
2864
2865 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
2866                                 u64 remote_handle, u64 ioba, u64 num_entries)
2867 {
2868         unsigned int ua = adapter->vdev->unit_address;
2869         struct device *dev = &adapter->vdev->dev;
2870         int rc;
2871
2872         /* Make sure the hypervisor sees the complete request */
2873         mb();
2874         rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
2875                                 cpu_to_be64(remote_handle),
2876                                 ioba, num_entries);
2877
2878         if (rc) {
2879                 if (rc == H_CLOSED)
2880                         dev_warn(dev, "CRQ Queue closed\n");
2881                 dev_err(dev, "Send (indirect) error (rc=%d)\n", rc);
2882         }
2883
2884         return rc;
2885 }
2886
2887 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
2888                             union ibmvnic_crq *crq)
2889 {
2890         unsigned int ua = adapter->vdev->unit_address;
2891         struct device *dev = &adapter->vdev->dev;
2892         u64 *u64_crq = (u64 *)crq;
2893         int rc;
2894
2895         netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
2896                    (unsigned long int)cpu_to_be64(u64_crq[0]),
2897                    (unsigned long int)cpu_to_be64(u64_crq[1]));
2898
2899         /* Make sure the hypervisor sees the complete request */
2900         mb();
2901
2902         rc = plpar_hcall_norets(H_SEND_CRQ, ua,
2903                                 cpu_to_be64(u64_crq[0]),
2904                                 cpu_to_be64(u64_crq[1]));
2905
2906         if (rc) {
2907                 if (rc == H_CLOSED)
2908                         dev_warn(dev, "CRQ Queue closed\n");
2909                 dev_warn(dev, "Send error (rc=%d)\n", rc);
2910         }
2911
2912         return rc;
2913 }
2914
2915 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
2916 {
2917         union ibmvnic_crq crq;
2918
2919         memset(&crq, 0, sizeof(crq));
2920         crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
2921         crq.generic.cmd = IBMVNIC_CRQ_INIT;
2922         netdev_dbg(adapter->netdev, "Sending CRQ init\n");
2923
2924         return ibmvnic_send_crq(adapter, &crq);
2925 }
2926
2927 static int send_version_xchg(struct ibmvnic_adapter *adapter)
2928 {
2929         union ibmvnic_crq crq;
2930
2931         memset(&crq, 0, sizeof(crq));
2932         crq.version_exchange.first = IBMVNIC_CRQ_CMD;
2933         crq.version_exchange.cmd = VERSION_EXCHANGE;
2934         crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
2935
2936         return ibmvnic_send_crq(adapter, &crq);
2937 }
2938
2939 struct vnic_login_client_data {
2940         u8      type;
2941         __be16  len;
2942         char    name;
2943 } __packed;
2944
2945 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
2946 {
2947         int len;
2948
2949         /* Calculate the amount of buffer space needed for the
2950          * vnic client data in the login buffer. There are four entries,
2951          * OS name, LPAR name, device name, and a null last entry.
2952          */
2953         len = 4 * sizeof(struct vnic_login_client_data);
2954         len += 6; /* "Linux" plus NULL */
2955         len += strlen(utsname()->nodename) + 1;
2956         len += strlen(adapter->netdev->name) + 1;
2957
2958         return len;
2959 }
2960
2961 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
2962                                  struct vnic_login_client_data *vlcd)
2963 {
2964         const char *os_name = "Linux";
2965         int len;
2966
2967         /* Type 1 - LPAR OS */
2968         vlcd->type = 1;
2969         len = strlen(os_name) + 1;
2970         vlcd->len = cpu_to_be16(len);
2971         strncpy(&vlcd->name, os_name, len);
2972         vlcd = (struct vnic_login_client_data *)((char *)&vlcd->name + len);
2973
2974         /* Type 2 - LPAR name */
2975         vlcd->type = 2;
2976         len = strlen(utsname()->nodename) + 1;
2977         vlcd->len = cpu_to_be16(len);
2978         strncpy(&vlcd->name, utsname()->nodename, len);
2979         vlcd = (struct vnic_login_client_data *)((char *)&vlcd->name + len);
2980
2981         /* Type 3 - device name */
2982         vlcd->type = 3;
2983         len = strlen(adapter->netdev->name) + 1;
2984         vlcd->len = cpu_to_be16(len);
2985         strncpy(&vlcd->name, adapter->netdev->name, len);
2986 }
2987
2988 static void send_login(struct ibmvnic_adapter *adapter)
2989 {
2990         struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
2991         struct ibmvnic_login_buffer *login_buffer;
2992         struct device *dev = &adapter->vdev->dev;
2993         dma_addr_t rsp_buffer_token;
2994         dma_addr_t buffer_token;
2995         size_t rsp_buffer_size;
2996         union ibmvnic_crq crq;
2997         size_t buffer_size;
2998         __be64 *tx_list_p;
2999         __be64 *rx_list_p;
3000         int client_data_len;
3001         struct vnic_login_client_data *vlcd;
3002         int i;
3003
3004         client_data_len = vnic_client_data_len(adapter);
3005
3006         buffer_size =
3007             sizeof(struct ibmvnic_login_buffer) +
3008             sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3009             client_data_len;
3010
3011         login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3012         if (!login_buffer)
3013                 goto buf_alloc_failed;
3014
3015         buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3016                                       DMA_TO_DEVICE);
3017         if (dma_mapping_error(dev, buffer_token)) {
3018                 dev_err(dev, "Couldn't map login buffer\n");
3019                 goto buf_map_failed;
3020         }
3021
3022         rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3023                           sizeof(u64) * adapter->req_tx_queues +
3024                           sizeof(u64) * adapter->req_rx_queues +
3025                           sizeof(u64) * adapter->req_rx_queues +
3026                           sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3027
3028         login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3029         if (!login_rsp_buffer)
3030                 goto buf_rsp_alloc_failed;
3031
3032         rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3033                                           rsp_buffer_size, DMA_FROM_DEVICE);
3034         if (dma_mapping_error(dev, rsp_buffer_token)) {
3035                 dev_err(dev, "Couldn't map login rsp buffer\n");
3036                 goto buf_rsp_map_failed;
3037         }
3038
3039         adapter->login_buf = login_buffer;
3040         adapter->login_buf_token = buffer_token;
3041         adapter->login_buf_sz = buffer_size;
3042         adapter->login_rsp_buf = login_rsp_buffer;
3043         adapter->login_rsp_buf_token = rsp_buffer_token;
3044         adapter->login_rsp_buf_sz = rsp_buffer_size;
3045
3046         login_buffer->len = cpu_to_be32(buffer_size);
3047         login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3048         login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3049         login_buffer->off_txcomp_subcrqs =
3050             cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3051         login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3052         login_buffer->off_rxcomp_subcrqs =
3053             cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3054                         sizeof(u64) * adapter->req_tx_queues);
3055         login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3056         login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3057
3058         tx_list_p = (__be64 *)((char *)login_buffer +
3059                                       sizeof(struct ibmvnic_login_buffer));
3060         rx_list_p = (__be64 *)((char *)login_buffer +
3061                                       sizeof(struct ibmvnic_login_buffer) +
3062                                       sizeof(u64) * adapter->req_tx_queues);
3063
3064         for (i = 0; i < adapter->req_tx_queues; i++) {
3065                 if (adapter->tx_scrq[i]) {
3066                         tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3067                                                    crq_num);
3068                 }
3069         }
3070
3071         for (i = 0; i < adapter->req_rx_queues; i++) {
3072                 if (adapter->rx_scrq[i]) {
3073                         rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3074                                                    crq_num);
3075                 }
3076         }
3077
3078         /* Insert vNIC login client data */
3079         vlcd = (struct vnic_login_client_data *)
3080                 ((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3081         login_buffer->client_data_offset =
3082                         cpu_to_be32((char *)vlcd - (char *)login_buffer);
3083         login_buffer->client_data_len = cpu_to_be32(client_data_len);
3084
3085         vnic_add_client_data(adapter, vlcd);
3086
3087         netdev_dbg(adapter->netdev, "Login Buffer:\n");
3088         for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3089                 netdev_dbg(adapter->netdev, "%016lx\n",
3090                            ((unsigned long int *)(adapter->login_buf))[i]);
3091         }
3092
3093         memset(&crq, 0, sizeof(crq));
3094         crq.login.first = IBMVNIC_CRQ_CMD;
3095         crq.login.cmd = LOGIN;
3096         crq.login.ioba = cpu_to_be32(buffer_token);
3097         crq.login.len = cpu_to_be32(buffer_size);
3098         ibmvnic_send_crq(adapter, &crq);
3099
3100         return;
3101
3102 buf_rsp_map_failed:
3103         kfree(login_rsp_buffer);
3104 buf_rsp_alloc_failed:
3105         dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3106 buf_map_failed:
3107         kfree(login_buffer);
3108 buf_alloc_failed:
3109         return;
3110 }
3111
3112 static void send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3113                              u32 len, u8 map_id)
3114 {
3115         union ibmvnic_crq crq;
3116
3117         memset(&crq, 0, sizeof(crq));
3118         crq.request_map.first = IBMVNIC_CRQ_CMD;
3119         crq.request_map.cmd = REQUEST_MAP;
3120         crq.request_map.map_id = map_id;
3121         crq.request_map.ioba = cpu_to_be32(addr);
3122         crq.request_map.len = cpu_to_be32(len);
3123         ibmvnic_send_crq(adapter, &crq);
3124 }
3125
3126 static void send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3127 {
3128         union ibmvnic_crq crq;
3129
3130         memset(&crq, 0, sizeof(crq));
3131         crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3132         crq.request_unmap.cmd = REQUEST_UNMAP;
3133         crq.request_unmap.map_id = map_id;
3134         ibmvnic_send_crq(adapter, &crq);
3135 }
3136
3137 static void send_map_query(struct ibmvnic_adapter *adapter)
3138 {
3139         union ibmvnic_crq crq;
3140
3141         memset(&crq, 0, sizeof(crq));
3142         crq.query_map.first = IBMVNIC_CRQ_CMD;
3143         crq.query_map.cmd = QUERY_MAP;
3144         ibmvnic_send_crq(adapter, &crq);
3145 }
3146
3147 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3148 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3149 {
3150         union ibmvnic_crq crq;
3151
3152         atomic_set(&adapter->running_cap_crqs, 0);
3153         memset(&crq, 0, sizeof(crq));
3154         crq.query_capability.first = IBMVNIC_CRQ_CMD;
3155         crq.query_capability.cmd = QUERY_CAPABILITY;
3156
3157         crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3158         atomic_inc(&adapter->running_cap_crqs);
3159         ibmvnic_send_crq(adapter, &crq);
3160
3161         crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3162         atomic_inc(&adapter->running_cap_crqs);
3163         ibmvnic_send_crq(adapter, &crq);
3164
3165         crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3166         atomic_inc(&adapter->running_cap_crqs);
3167         ibmvnic_send_crq(adapter, &crq);
3168
3169         crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3170         atomic_inc(&adapter->running_cap_crqs);
3171         ibmvnic_send_crq(adapter, &crq);
3172
3173         crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3174         atomic_inc(&adapter->running_cap_crqs);
3175         ibmvnic_send_crq(adapter, &crq);
3176
3177         crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3178         atomic_inc(&adapter->running_cap_crqs);
3179         ibmvnic_send_crq(adapter, &crq);
3180
3181         crq.query_capability.capability =
3182             cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3183         atomic_inc(&adapter->running_cap_crqs);
3184         ibmvnic_send_crq(adapter, &crq);
3185
3186         crq.query_capability.capability =
3187             cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3188         atomic_inc(&adapter->running_cap_crqs);
3189         ibmvnic_send_crq(adapter, &crq);
3190
3191         crq.query_capability.capability =
3192             cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3193         atomic_inc(&adapter->running_cap_crqs);
3194         ibmvnic_send_crq(adapter, &crq);
3195
3196         crq.query_capability.capability =
3197             cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3198         atomic_inc(&adapter->running_cap_crqs);
3199         ibmvnic_send_crq(adapter, &crq);
3200
3201         crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3202         atomic_inc(&adapter->running_cap_crqs);
3203         ibmvnic_send_crq(adapter, &crq);
3204
3205         crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3206         atomic_inc(&adapter->running_cap_crqs);
3207         ibmvnic_send_crq(adapter, &crq);
3208
3209         crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3210         atomic_inc(&adapter->running_cap_crqs);
3211         ibmvnic_send_crq(adapter, &crq);
3212
3213         crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3214         atomic_inc(&adapter->running_cap_crqs);
3215         ibmvnic_send_crq(adapter, &crq);
3216
3217         crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3218         atomic_inc(&adapter->running_cap_crqs);
3219         ibmvnic_send_crq(adapter, &crq);
3220
3221         crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3222         atomic_inc(&adapter->running_cap_crqs);
3223         ibmvnic_send_crq(adapter, &crq);
3224
3225         crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3226         atomic_inc(&adapter->running_cap_crqs);
3227         ibmvnic_send_crq(adapter, &crq);
3228
3229         crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3230         atomic_inc(&adapter->running_cap_crqs);
3231         ibmvnic_send_crq(adapter, &crq);
3232
3233         crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3234         atomic_inc(&adapter->running_cap_crqs);
3235         ibmvnic_send_crq(adapter, &crq);
3236
3237         crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3238         atomic_inc(&adapter->running_cap_crqs);
3239         ibmvnic_send_crq(adapter, &crq);
3240
3241         crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3242         atomic_inc(&adapter->running_cap_crqs);
3243         ibmvnic_send_crq(adapter, &crq);
3244
3245         crq.query_capability.capability =
3246                         cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3247         atomic_inc(&adapter->running_cap_crqs);
3248         ibmvnic_send_crq(adapter, &crq);
3249
3250         crq.query_capability.capability =
3251                         cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3252         atomic_inc(&adapter->running_cap_crqs);
3253         ibmvnic_send_crq(adapter, &crq);
3254
3255         crq.query_capability.capability =
3256                         cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3257         atomic_inc(&adapter->running_cap_crqs);
3258         ibmvnic_send_crq(adapter, &crq);
3259
3260         crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3261         atomic_inc(&adapter->running_cap_crqs);
3262         ibmvnic_send_crq(adapter, &crq);
3263 }
3264
3265 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3266                                 struct ibmvnic_adapter *adapter)
3267 {
3268         struct device *dev = &adapter->vdev->dev;
3269
3270         if (crq->get_vpd_size_rsp.rc.code) {
3271                 dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3272                         crq->get_vpd_size_rsp.rc.code);
3273                 complete(&adapter->fw_done);
3274                 return;
3275         }
3276
3277         adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3278         complete(&adapter->fw_done);
3279 }
3280
3281 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3282                            struct ibmvnic_adapter *adapter)
3283 {
3284         struct device *dev = &adapter->vdev->dev;
3285         unsigned char *substr = NULL, *ptr = NULL;
3286         u8 fw_level_len = 0;
3287
3288         memset(adapter->fw_version, 0, 32);
3289
3290         dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3291                          DMA_FROM_DEVICE);
3292
3293         if (crq->get_vpd_rsp.rc.code) {
3294                 dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3295                         crq->get_vpd_rsp.rc.code);
3296                 goto complete;
3297         }
3298
3299         /* get the position of the firmware version info
3300          * located after the ASCII 'RM' substring in the buffer
3301          */
3302         substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3303         if (!substr) {
3304                 dev_info(dev, "No FW level provided by VPD\n");
3305                 goto complete;
3306         }
3307
3308         /* get length of firmware level ASCII substring */
3309         if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3310                 fw_level_len = *(substr + 2);
3311         } else {
3312                 dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3313                 goto complete;
3314         }
3315
3316         /* copy firmware version string from vpd into adapter */
3317         if ((substr + 3 + fw_level_len) <
3318             (adapter->vpd->buff + adapter->vpd->len)) {
3319                 ptr = strncpy((char *)adapter->fw_version,
3320                               substr + 3, fw_level_len);
3321
3322                 if (!ptr)
3323                         dev_err(dev, "Failed to isolate FW level string\n");
3324         } else {
3325                 dev_info(dev, "FW substr extrapolated VPD buff\n");
3326         }
3327
3328 complete:
3329         complete(&adapter->fw_done);
3330 }
3331
3332 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3333 {
3334         struct device *dev = &adapter->vdev->dev;
3335         struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3336         union ibmvnic_crq crq;
3337         int i;
3338
3339         dma_unmap_single(dev, adapter->ip_offload_tok,
3340                          sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
3341
3342         netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
3343         for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
3344                 netdev_dbg(adapter->netdev, "%016lx\n",
3345                            ((unsigned long int *)(buf))[i]);
3346
3347         netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
3348         netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
3349         netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
3350                    buf->tcp_ipv4_chksum);
3351         netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
3352                    buf->tcp_ipv6_chksum);
3353         netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
3354                    buf->udp_ipv4_chksum);
3355         netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
3356                    buf->udp_ipv6_chksum);
3357         netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
3358                    buf->large_tx_ipv4);
3359         netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
3360                    buf->large_tx_ipv6);
3361         netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
3362                    buf->large_rx_ipv4);
3363         netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
3364                    buf->large_rx_ipv6);
3365         netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
3366                    buf->max_ipv4_header_size);
3367         netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
3368                    buf->max_ipv6_header_size);
3369         netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
3370                    buf->max_tcp_header_size);
3371         netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
3372                    buf->max_udp_header_size);
3373         netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
3374                    buf->max_large_tx_size);
3375         netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
3376                    buf->max_large_rx_size);
3377         netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
3378                    buf->ipv6_extension_header);
3379         netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
3380                    buf->tcp_pseudosum_req);
3381         netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
3382                    buf->num_ipv6_ext_headers);
3383         netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
3384                    buf->off_ipv6_ext_headers);
3385
3386         adapter->ip_offload_ctrl_tok =
3387             dma_map_single(dev, &adapter->ip_offload_ctrl,
3388                            sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
3389
3390         if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
3391                 dev_err(dev, "Couldn't map ip offload control buffer\n");
3392                 return;
3393         }
3394
3395         adapter->ip_offload_ctrl.len =
3396             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3397         adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
3398         adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
3399         adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
3400         adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
3401         adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
3402         adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
3403         adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
3404         adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
3405         adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
3406
3407         /* large_rx disabled for now, additional features needed */
3408         adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
3409         adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
3410
3411         adapter->netdev->features = NETIF_F_SG | NETIF_F_GSO;
3412
3413         if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
3414                 adapter->netdev->features |= NETIF_F_IP_CSUM;
3415
3416         if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
3417                 adapter->netdev->features |= NETIF_F_IPV6_CSUM;
3418
3419         if ((adapter->netdev->features &
3420             (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
3421                 adapter->netdev->features |= NETIF_F_RXCSUM;
3422
3423         if (buf->large_tx_ipv4)
3424                 adapter->netdev->features |= NETIF_F_TSO;
3425         if (buf->large_tx_ipv6)
3426                 adapter->netdev->features |= NETIF_F_TSO6;
3427
3428         adapter->netdev->hw_features |= adapter->netdev->features;
3429
3430         memset(&crq, 0, sizeof(crq));
3431         crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
3432         crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
3433         crq.control_ip_offload.len =
3434             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3435         crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
3436         ibmvnic_send_crq(adapter, &crq);
3437 }
3438
3439 static void handle_error_info_rsp(union ibmvnic_crq *crq,
3440                                   struct ibmvnic_adapter *adapter)
3441 {
3442         struct device *dev = &adapter->vdev->dev;
3443         struct ibmvnic_error_buff *error_buff, *tmp;
3444         unsigned long flags;
3445         bool found = false;
3446         int i;
3447
3448         if (!crq->request_error_rsp.rc.code) {
3449                 dev_info(dev, "Request Error Rsp returned with rc=%x\n",
3450                          crq->request_error_rsp.rc.code);
3451                 return;
3452         }
3453
3454         spin_lock_irqsave(&adapter->error_list_lock, flags);
3455         list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list)
3456                 if (error_buff->error_id == crq->request_error_rsp.error_id) {
3457                         found = true;
3458                         list_del(&error_buff->list);
3459                         break;
3460                 }
3461         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3462
3463         if (!found) {
3464                 dev_err(dev, "Couldn't find error id %x\n",
3465                         be32_to_cpu(crq->request_error_rsp.error_id));
3466                 return;
3467         }
3468
3469         dev_err(dev, "Detailed info for error id %x:",
3470                 be32_to_cpu(crq->request_error_rsp.error_id));
3471
3472         for (i = 0; i < error_buff->len; i++) {
3473                 pr_cont("%02x", (int)error_buff->buff[i]);
3474                 if (i % 8 == 7)
3475                         pr_cont(" ");
3476         }
3477         pr_cont("\n");
3478
3479         dma_unmap_single(dev, error_buff->dma, error_buff->len,
3480                          DMA_FROM_DEVICE);
3481         kfree(error_buff->buff);
3482         kfree(error_buff);
3483 }
3484
3485 static void request_error_information(struct ibmvnic_adapter *adapter,
3486                                       union ibmvnic_crq *err_crq)
3487 {
3488         struct device *dev = &adapter->vdev->dev;
3489         struct net_device *netdev = adapter->netdev;
3490         struct ibmvnic_error_buff *error_buff;
3491         unsigned long timeout = msecs_to_jiffies(30000);
3492         union ibmvnic_crq crq;
3493         unsigned long flags;
3494         int rc, detail_len;
3495
3496         error_buff = kmalloc(sizeof(*error_buff), GFP_ATOMIC);
3497         if (!error_buff)
3498                 return;
3499
3500         detail_len = be32_to_cpu(err_crq->error_indication.detail_error_sz);
3501         error_buff->buff = kmalloc(detail_len, GFP_ATOMIC);
3502         if (!error_buff->buff) {
3503                 kfree(error_buff);
3504                 return;
3505         }
3506
3507         error_buff->dma = dma_map_single(dev, error_buff->buff, detail_len,
3508                                          DMA_FROM_DEVICE);
3509         if (dma_mapping_error(dev, error_buff->dma)) {
3510                 netdev_err(netdev, "Couldn't map error buffer\n");
3511                 kfree(error_buff->buff);
3512                 kfree(error_buff);
3513                 return;
3514         }
3515
3516         error_buff->len = detail_len;
3517         error_buff->error_id = err_crq->error_indication.error_id;
3518
3519         spin_lock_irqsave(&adapter->error_list_lock, flags);
3520         list_add_tail(&error_buff->list, &adapter->errors);
3521         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3522
3523         memset(&crq, 0, sizeof(crq));
3524         crq.request_error_info.first = IBMVNIC_CRQ_CMD;
3525         crq.request_error_info.cmd = REQUEST_ERROR_INFO;
3526         crq.request_error_info.ioba = cpu_to_be32(error_buff->dma);
3527         crq.request_error_info.len = cpu_to_be32(detail_len);
3528         crq.request_error_info.error_id = err_crq->error_indication.error_id;
3529
3530         rc = ibmvnic_send_crq(adapter, &crq);
3531         if (rc) {
3532                 netdev_err(netdev, "failed to request error information\n");
3533                 goto err_info_fail;
3534         }
3535
3536         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
3537                 netdev_err(netdev, "timeout waiting for error information\n");
3538                 goto err_info_fail;
3539         }
3540
3541         return;
3542
3543 err_info_fail:
3544         spin_lock_irqsave(&adapter->error_list_lock, flags);
3545         list_del(&error_buff->list);
3546         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3547
3548         kfree(error_buff->buff);
3549         kfree(error_buff);
3550 }
3551
3552 static void handle_error_indication(union ibmvnic_crq *crq,
3553                                     struct ibmvnic_adapter *adapter)
3554 {
3555         struct device *dev = &adapter->vdev->dev;
3556
3557         dev_err(dev, "Firmware reports %serror id %x, cause %d\n",
3558                 crq->error_indication.flags
3559                         & IBMVNIC_FATAL_ERROR ? "FATAL " : "",
3560                 be32_to_cpu(crq->error_indication.error_id),
3561                 be16_to_cpu(crq->error_indication.error_cause));
3562
3563         if (be32_to_cpu(crq->error_indication.error_id))
3564                 request_error_information(adapter, crq);
3565
3566         if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
3567                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3568         else
3569                 ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
3570 }
3571
3572 static void handle_change_mac_rsp(union ibmvnic_crq *crq,
3573                                   struct ibmvnic_adapter *adapter)
3574 {
3575         struct net_device *netdev = adapter->netdev;
3576         struct device *dev = &adapter->vdev->dev;
3577         long rc;
3578
3579         rc = crq->change_mac_addr_rsp.rc.code;
3580         if (rc) {
3581                 dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
3582                 return;
3583         }
3584         memcpy(netdev->dev_addr, &crq->change_mac_addr_rsp.mac_addr[0],
3585                ETH_ALEN);
3586 }
3587
3588 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
3589                                    struct ibmvnic_adapter *adapter)
3590 {
3591         struct device *dev = &adapter->vdev->dev;
3592         u64 *req_value;
3593         char *name;
3594
3595         atomic_dec(&adapter->running_cap_crqs);
3596         switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
3597         case REQ_TX_QUEUES:
3598                 req_value = &adapter->req_tx_queues;
3599                 name = "tx";
3600                 break;
3601         case REQ_RX_QUEUES:
3602                 req_value = &adapter->req_rx_queues;
3603                 name = "rx";
3604                 break;
3605         case REQ_RX_ADD_QUEUES:
3606                 req_value = &adapter->req_rx_add_queues;
3607                 name = "rx_add";
3608                 break;
3609         case REQ_TX_ENTRIES_PER_SUBCRQ:
3610                 req_value = &adapter->req_tx_entries_per_subcrq;
3611                 name = "tx_entries_per_subcrq";
3612                 break;
3613         case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
3614                 req_value = &adapter->req_rx_add_entries_per_subcrq;
3615                 name = "rx_add_entries_per_subcrq";
3616                 break;
3617         case REQ_MTU:
3618                 req_value = &adapter->req_mtu;
3619                 name = "mtu";
3620                 break;
3621         case PROMISC_REQUESTED:
3622                 req_value = &adapter->promisc;
3623                 name = "promisc";
3624                 break;
3625         default:
3626                 dev_err(dev, "Got invalid cap request rsp %d\n",
3627                         crq->request_capability.capability);
3628                 return;
3629         }
3630
3631         switch (crq->request_capability_rsp.rc.code) {
3632         case SUCCESS:
3633                 break;
3634         case PARTIALSUCCESS:
3635                 dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3636                          *req_value,
3637                          (long int)be64_to_cpu(crq->request_capability_rsp.
3638                                                number), name);
3639
3640                 if (be16_to_cpu(crq->request_capability_rsp.capability) ==
3641                     REQ_MTU) {
3642                         pr_err("mtu of %llu is not supported. Reverting.\n",
3643                                *req_value);
3644                         *req_value = adapter->fallback.mtu;
3645                 } else {
3646                         *req_value =
3647                                 be64_to_cpu(crq->request_capability_rsp.number);
3648                 }
3649
3650                 ibmvnic_send_req_caps(adapter, 1);
3651                 return;
3652         default:
3653                 dev_err(dev, "Error %d in request cap rsp\n",
3654                         crq->request_capability_rsp.rc.code);
3655                 return;
3656         }
3657
3658         /* Done receiving requested capabilities, query IP offload support */
3659         if (atomic_read(&adapter->running_cap_crqs) == 0) {
3660                 union ibmvnic_crq newcrq;
3661                 int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
3662                 struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
3663                     &adapter->ip_offload_buf;
3664
3665                 adapter->wait_capability = false;
3666                 adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
3667                                                          buf_sz,
3668                                                          DMA_FROM_DEVICE);
3669
3670                 if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
3671                         if (!firmware_has_feature(FW_FEATURE_CMO))
3672                                 dev_err(dev, "Couldn't map offload buffer\n");
3673                         return;
3674                 }
3675
3676                 memset(&newcrq, 0, sizeof(newcrq));
3677                 newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
3678                 newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
3679                 newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
3680                 newcrq.query_ip_offload.ioba =
3681                     cpu_to_be32(adapter->ip_offload_tok);
3682
3683                 ibmvnic_send_crq(adapter, &newcrq);
3684         }
3685 }
3686
3687 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
3688                             struct ibmvnic_adapter *adapter)
3689 {
3690         struct device *dev = &adapter->vdev->dev;
3691         struct net_device *netdev = adapter->netdev;
3692         struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
3693         struct ibmvnic_login_buffer *login = adapter->login_buf;
3694         int i;
3695
3696         dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
3697                          DMA_BIDIRECTIONAL);
3698         dma_unmap_single(dev, adapter->login_rsp_buf_token,
3699                          adapter->login_rsp_buf_sz, DMA_BIDIRECTIONAL);
3700
3701         /* If the number of queues requested can't be allocated by the
3702          * server, the login response will return with code 1. We will need
3703          * to resend the login buffer with fewer queues requested.
3704          */
3705         if (login_rsp_crq->generic.rc.code) {
3706                 adapter->renegotiate = true;
3707                 complete(&adapter->init_done);
3708                 return 0;
3709         }
3710
3711         netdev->mtu = adapter->req_mtu - ETH_HLEN;
3712
3713         netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
3714         for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
3715                 netdev_dbg(adapter->netdev, "%016lx\n",
3716                            ((unsigned long int *)(adapter->login_rsp_buf))[i]);
3717         }
3718
3719         /* Sanity checks */
3720         if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
3721             (be32_to_cpu(login->num_rxcomp_subcrqs) *
3722              adapter->req_rx_add_queues !=
3723              be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
3724                 dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
3725                 ibmvnic_remove(adapter->vdev);
3726                 return -EIO;
3727         }
3728         complete(&adapter->init_done);
3729
3730         return 0;
3731 }
3732
3733 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
3734                                      struct ibmvnic_adapter *adapter)
3735 {
3736         struct device *dev = &adapter->vdev->dev;
3737         long rc;
3738
3739         rc = crq->request_unmap_rsp.rc.code;
3740         if (rc)
3741                 dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
3742 }
3743
3744 static void handle_query_map_rsp(union ibmvnic_crq *crq,
3745                                  struct ibmvnic_adapter *adapter)
3746 {
3747         struct net_device *netdev = adapter->netdev;
3748         struct device *dev = &adapter->vdev->dev;
3749         long rc;
3750
3751         rc = crq->query_map_rsp.rc.code;
3752         if (rc) {
3753                 dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
3754                 return;
3755         }
3756         netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
3757                    crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
3758                    crq->query_map_rsp.free_pages);
3759 }
3760
3761 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
3762                                  struct ibmvnic_adapter *adapter)
3763 {
3764         struct net_device *netdev = adapter->netdev;
3765         struct device *dev = &adapter->vdev->dev;
3766         long rc;
3767
3768         atomic_dec(&adapter->running_cap_crqs);
3769         netdev_dbg(netdev, "Outstanding queries: %d\n",
3770                    atomic_read(&adapter->running_cap_crqs));
3771         rc = crq->query_capability.rc.code;
3772         if (rc) {
3773                 dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
3774                 goto out;
3775         }
3776
3777         switch (be16_to_cpu(crq->query_capability.capability)) {
3778         case MIN_TX_QUEUES:
3779                 adapter->min_tx_queues =
3780                     be64_to_cpu(crq->query_capability.number);
3781                 netdev_dbg(netdev, "min_tx_queues = %lld\n",
3782                            adapter->min_tx_queues);
3783                 break;
3784         case MIN_RX_QUEUES:
3785                 adapter->min_rx_queues =
3786                     be64_to_cpu(crq->query_capability.number);
3787                 netdev_dbg(netdev, "min_rx_queues = %lld\n",
3788                            adapter->min_rx_queues);
3789                 break;
3790         case MIN_RX_ADD_QUEUES:
3791                 adapter->min_rx_add_queues =
3792                     be64_to_cpu(crq->query_capability.number);
3793                 netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
3794                            adapter->min_rx_add_queues);
3795                 break;
3796         case MAX_TX_QUEUES:
3797                 adapter->max_tx_queues =
3798                     be64_to_cpu(crq->query_capability.number);
3799                 netdev_dbg(netdev, "max_tx_queues = %lld\n",
3800                            adapter->max_tx_queues);
3801                 break;
3802         case MAX_RX_QUEUES:
3803                 adapter->max_rx_queues =
3804                     be64_to_cpu(crq->query_capability.number);
3805                 netdev_dbg(netdev, "max_rx_queues = %lld\n",
3806                            adapter->max_rx_queues);
3807                 break;
3808         case MAX_RX_ADD_QUEUES:
3809                 adapter->max_rx_add_queues =
3810                     be64_to_cpu(crq->query_capability.number);
3811                 netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
3812                            adapter->max_rx_add_queues);
3813                 break;
3814         case MIN_TX_ENTRIES_PER_SUBCRQ:
3815                 adapter->min_tx_entries_per_subcrq =
3816                     be64_to_cpu(crq->query_capability.number);
3817                 netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
3818                            adapter->min_tx_entries_per_subcrq);
3819                 break;
3820         case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
3821                 adapter->min_rx_add_entries_per_subcrq =
3822                     be64_to_cpu(crq->query_capability.number);
3823                 netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
3824                            adapter->min_rx_add_entries_per_subcrq);
3825                 break;
3826         case MAX_TX_ENTRIES_PER_SUBCRQ:
3827                 adapter->max_tx_entries_per_subcrq =
3828                     be64_to_cpu(crq->query_capability.number);
3829                 netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
3830                            adapter->max_tx_entries_per_subcrq);
3831                 break;
3832         case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
3833                 adapter->max_rx_add_entries_per_subcrq =
3834                     be64_to_cpu(crq->query_capability.number);
3835                 netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
3836                            adapter->max_rx_add_entries_per_subcrq);
3837                 break;
3838         case TCP_IP_OFFLOAD:
3839                 adapter->tcp_ip_offload =
3840                     be64_to_cpu(crq->query_capability.number);
3841                 netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
3842                            adapter->tcp_ip_offload);
3843                 break;
3844         case PROMISC_SUPPORTED:
3845                 adapter->promisc_supported =
3846                     be64_to_cpu(crq->query_capability.number);
3847                 netdev_dbg(netdev, "promisc_supported = %lld\n",
3848                            adapter->promisc_supported);
3849                 break;
3850         case MIN_MTU:
3851                 adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
3852                 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
3853                 netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
3854                 break;
3855         case MAX_MTU:
3856                 adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
3857                 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
3858                 netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
3859                 break;
3860         case MAX_MULTICAST_FILTERS:
3861                 adapter->max_multicast_filters =
3862                     be64_to_cpu(crq->query_capability.number);
3863                 netdev_dbg(netdev, "max_multicast_filters = %lld\n",
3864                            adapter->max_multicast_filters);
3865                 break;
3866         case VLAN_HEADER_INSERTION:
3867                 adapter->vlan_header_insertion =
3868                     be64_to_cpu(crq->query_capability.number);
3869                 if (adapter->vlan_header_insertion)
3870                         netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
3871                 netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
3872                            adapter->vlan_header_insertion);
3873                 break;
3874         case RX_VLAN_HEADER_INSERTION:
3875                 adapter->rx_vlan_header_insertion =
3876                     be64_to_cpu(crq->query_capability.number);
3877                 netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
3878                            adapter->rx_vlan_header_insertion);
3879                 break;
3880         case MAX_TX_SG_ENTRIES:
3881                 adapter->max_tx_sg_entries =
3882                     be64_to_cpu(crq->query_capability.number);
3883                 netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
3884                            adapter->max_tx_sg_entries);
3885                 break;
3886         case RX_SG_SUPPORTED:
3887                 adapter->rx_sg_supported =
3888                     be64_to_cpu(crq->query_capability.number);
3889                 netdev_dbg(netdev, "rx_sg_supported = %lld\n",
3890                            adapter->rx_sg_supported);
3891                 break;
3892         case OPT_TX_COMP_SUB_QUEUES:
3893                 adapter->opt_tx_comp_sub_queues =
3894                     be64_to_cpu(crq->query_capability.number);
3895                 netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
3896                            adapter->opt_tx_comp_sub_queues);
3897                 break;
3898         case OPT_RX_COMP_QUEUES:
3899                 adapter->opt_rx_comp_queues =
3900                     be64_to_cpu(crq->query_capability.number);
3901                 netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
3902                            adapter->opt_rx_comp_queues);
3903                 break;
3904         case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
3905                 adapter->opt_rx_bufadd_q_per_rx_comp_q =
3906                     be64_to_cpu(crq->query_capability.number);
3907                 netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
3908                            adapter->opt_rx_bufadd_q_per_rx_comp_q);
3909                 break;
3910         case OPT_TX_ENTRIES_PER_SUBCRQ:
3911                 adapter->opt_tx_entries_per_subcrq =
3912                     be64_to_cpu(crq->query_capability.number);
3913                 netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
3914                            adapter->opt_tx_entries_per_subcrq);
3915                 break;
3916         case OPT_RXBA_ENTRIES_PER_SUBCRQ:
3917                 adapter->opt_rxba_entries_per_subcrq =
3918                     be64_to_cpu(crq->query_capability.number);
3919                 netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
3920                            adapter->opt_rxba_entries_per_subcrq);
3921                 break;
3922         case TX_RX_DESC_REQ:
3923                 adapter->tx_rx_desc_req = crq->query_capability.number;
3924                 netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
3925                            adapter->tx_rx_desc_req);
3926                 break;
3927
3928         default:
3929                 netdev_err(netdev, "Got invalid cap rsp %d\n",
3930                            crq->query_capability.capability);
3931         }
3932
3933 out:
3934         if (atomic_read(&adapter->running_cap_crqs) == 0) {
3935                 adapter->wait_capability = false;
3936                 ibmvnic_send_req_caps(adapter, 0);
3937         }
3938 }
3939
3940 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
3941                                struct ibmvnic_adapter *adapter)
3942 {
3943         struct ibmvnic_generic_crq *gen_crq = &crq->generic;
3944         struct net_device *netdev = adapter->netdev;
3945         struct device *dev = &adapter->vdev->dev;
3946         u64 *u64_crq = (u64 *)crq;
3947         long rc;
3948
3949         netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
3950                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3951                    (unsigned long int)cpu_to_be64(u64_crq[1]));
3952         switch (gen_crq->first) {
3953         case IBMVNIC_CRQ_INIT_RSP:
3954                 switch (gen_crq->cmd) {
3955                 case IBMVNIC_CRQ_INIT:
3956                         dev_info(dev, "Partner initialized\n");
3957                         adapter->from_passive_init = true;
3958                         complete(&adapter->init_done);
3959                         break;
3960                 case IBMVNIC_CRQ_INIT_COMPLETE:
3961                         dev_info(dev, "Partner initialization complete\n");
3962                         send_version_xchg(adapter);
3963                         break;
3964                 default:
3965                         dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
3966                 }
3967                 return;
3968         case IBMVNIC_CRQ_XPORT_EVENT:
3969                 netif_carrier_off(netdev);
3970                 if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
3971                         dev_info(dev, "Migrated, re-enabling adapter\n");
3972                         ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
3973                 } else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
3974                         dev_info(dev, "Backing device failover detected\n");
3975                         ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
3976                 } else {
3977                         /* The adapter lost the connection */
3978                         dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
3979                                 gen_crq->cmd);
3980                         ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3981                 }
3982                 return;
3983         case IBMVNIC_CRQ_CMD_RSP:
3984                 break;
3985         default:
3986                 dev_err(dev, "Got an invalid msg type 0x%02x\n",
3987                         gen_crq->first);
3988                 return;
3989         }
3990
3991         switch (gen_crq->cmd) {
3992         case VERSION_EXCHANGE_RSP:
3993                 rc = crq->version_exchange_rsp.rc.code;
3994                 if (rc) {
3995                         dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
3996                         break;
3997                 }
3998                 dev_info(dev, "Partner protocol version is %d\n",
3999                          crq->version_exchange_rsp.version);
4000                 if (be16_to_cpu(crq->version_exchange_rsp.version) <
4001                     ibmvnic_version)
4002                         ibmvnic_version =
4003                             be16_to_cpu(crq->version_exchange_rsp.version);
4004                 send_cap_queries(adapter);
4005                 break;
4006         case QUERY_CAPABILITY_RSP:
4007                 handle_query_cap_rsp(crq, adapter);
4008                 break;
4009         case QUERY_MAP_RSP:
4010                 handle_query_map_rsp(crq, adapter);
4011                 break;
4012         case REQUEST_MAP_RSP:
4013                 adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4014                 complete(&adapter->fw_done);
4015                 break;
4016         case REQUEST_UNMAP_RSP:
4017                 handle_request_unmap_rsp(crq, adapter);
4018                 break;
4019         case REQUEST_CAPABILITY_RSP:
4020                 handle_request_cap_rsp(crq, adapter);
4021                 break;
4022         case LOGIN_RSP:
4023                 netdev_dbg(netdev, "Got Login Response\n");
4024                 handle_login_rsp(crq, adapter);
4025                 break;
4026         case LOGICAL_LINK_STATE_RSP:
4027                 netdev_dbg(netdev,
4028                            "Got Logical Link State Response, state: %d rc: %d\n",
4029                            crq->logical_link_state_rsp.link_state,
4030                            crq->logical_link_state_rsp.rc.code);
4031                 adapter->logical_link_state =
4032                     crq->logical_link_state_rsp.link_state;
4033                 adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4034                 complete(&adapter->init_done);
4035                 break;
4036         case LINK_STATE_INDICATION:
4037                 netdev_dbg(netdev, "Got Logical Link State Indication\n");
4038                 adapter->phys_link_state =
4039                     crq->link_state_indication.phys_link_state;
4040                 adapter->logical_link_state =
4041                     crq->link_state_indication.logical_link_state;
4042                 break;
4043         case CHANGE_MAC_ADDR_RSP:
4044                 netdev_dbg(netdev, "Got MAC address change Response\n");
4045                 handle_change_mac_rsp(crq, adapter);
4046                 break;
4047         case ERROR_INDICATION:
4048                 netdev_dbg(netdev, "Got Error Indication\n");
4049                 handle_error_indication(crq, adapter);
4050                 break;
4051         case REQUEST_ERROR_RSP:
4052                 netdev_dbg(netdev, "Got Error Detail Response\n");
4053                 handle_error_info_rsp(crq, adapter);
4054                 break;
4055         case REQUEST_STATISTICS_RSP:
4056                 netdev_dbg(netdev, "Got Statistics Response\n");
4057                 complete(&adapter->stats_done);
4058                 break;
4059         case QUERY_IP_OFFLOAD_RSP:
4060                 netdev_dbg(netdev, "Got Query IP offload Response\n");
4061                 handle_query_ip_offload_rsp(adapter);
4062                 break;
4063         case MULTICAST_CTRL_RSP:
4064                 netdev_dbg(netdev, "Got multicast control Response\n");
4065                 break;
4066         case CONTROL_IP_OFFLOAD_RSP:
4067                 netdev_dbg(netdev, "Got Control IP offload Response\n");
4068                 dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4069                                  sizeof(adapter->ip_offload_ctrl),
4070                                  DMA_TO_DEVICE);
4071                 complete(&adapter->init_done);
4072                 break;
4073         case COLLECT_FW_TRACE_RSP:
4074                 netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4075                 complete(&adapter->fw_done);
4076                 break;
4077         case GET_VPD_SIZE_RSP:
4078                 handle_vpd_size_rsp(crq, adapter);
4079                 break;
4080         case GET_VPD_RSP:
4081                 handle_vpd_rsp(crq, adapter);
4082                 break;
4083         default:
4084                 netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4085                            gen_crq->cmd);
4086         }
4087 }
4088
4089 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4090 {
4091         struct ibmvnic_adapter *adapter = instance;
4092
4093         tasklet_schedule(&adapter->tasklet);
4094         return IRQ_HANDLED;
4095 }
4096
4097 static void ibmvnic_tasklet(void *data)
4098 {
4099         struct ibmvnic_adapter *adapter = data;
4100         struct ibmvnic_crq_queue *queue = &adapter->crq;
4101         union ibmvnic_crq *crq;
4102         unsigned long flags;
4103         bool done = false;
4104
4105         spin_lock_irqsave(&queue->lock, flags);
4106         while (!done) {
4107                 /* Pull all the valid messages off the CRQ */
4108                 while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4109                         ibmvnic_handle_crq(crq, adapter);
4110                         crq->generic.first = 0;
4111                 }
4112
4113                 /* remain in tasklet until all
4114                  * capabilities responses are received
4115                  */
4116                 if (!adapter->wait_capability)
4117                         done = true;
4118         }
4119         /* if capabilities CRQ's were sent in this tasklet, the following
4120          * tasklet must wait until all responses are received
4121          */
4122         if (atomic_read(&adapter->running_cap_crqs) != 0)
4123                 adapter->wait_capability = true;
4124         spin_unlock_irqrestore(&queue->lock, flags);
4125 }
4126
4127 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4128 {
4129         struct vio_dev *vdev = adapter->vdev;
4130         int rc;
4131
4132         do {
4133                 rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4134         } while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4135
4136         if (rc)
4137                 dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4138
4139         return rc;
4140 }
4141
4142 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4143 {
4144         struct ibmvnic_crq_queue *crq = &adapter->crq;
4145         struct device *dev = &adapter->vdev->dev;
4146         struct vio_dev *vdev = adapter->vdev;
4147         int rc;
4148
4149         /* Close the CRQ */
4150         do {
4151                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4152         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4153
4154         /* Clean out the queue */
4155         memset(crq->msgs, 0, PAGE_SIZE);
4156         crq->cur = 0;
4157
4158         /* And re-open it again */
4159         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4160                                 crq->msg_token, PAGE_SIZE);
4161
4162         if (rc == H_CLOSED)
4163                 /* Adapter is good, but other end is not ready */
4164                 dev_warn(dev, "Partner adapter not ready\n");
4165         else if (rc != 0)
4166                 dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4167
4168         return rc;
4169 }
4170
4171 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4172 {
4173         struct ibmvnic_crq_queue *crq = &adapter->crq;
4174         struct vio_dev *vdev = adapter->vdev;
4175         long rc;
4176
4177         if (!crq->msgs)
4178                 return;
4179
4180         netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4181         free_irq(vdev->irq, adapter);
4182         tasklet_kill(&adapter->tasklet);
4183         do {
4184                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4185         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4186
4187         dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4188                          DMA_BIDIRECTIONAL);
4189         free_page((unsigned long)crq->msgs);
4190         crq->msgs = NULL;
4191 }
4192
4193 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4194 {
4195         struct ibmvnic_crq_queue *crq = &adapter->crq;
4196         struct device *dev = &adapter->vdev->dev;
4197         struct vio_dev *vdev = adapter->vdev;
4198         int rc, retrc = -ENOMEM;
4199
4200         if (crq->msgs)
4201                 return 0;
4202
4203         crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4204         /* Should we allocate more than one page? */
4205
4206         if (!crq->msgs)
4207                 return -ENOMEM;
4208
4209         crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4210         crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4211                                         DMA_BIDIRECTIONAL);
4212         if (dma_mapping_error(dev, crq->msg_token))
4213                 goto map_failed;
4214
4215         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4216                                 crq->msg_token, PAGE_SIZE);
4217
4218         if (rc == H_RESOURCE)
4219                 /* maybe kexecing and resource is busy. try a reset */
4220                 rc = ibmvnic_reset_crq(adapter);
4221         retrc = rc;
4222
4223         if (rc == H_CLOSED) {
4224                 dev_warn(dev, "Partner adapter not ready\n");
4225         } else if (rc) {
4226                 dev_warn(dev, "Error %d opening adapter\n", rc);
4227                 goto reg_crq_failed;
4228         }
4229
4230         retrc = 0;
4231
4232         tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4233                      (unsigned long)adapter);
4234
4235         netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4236         rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, IBMVNIC_NAME,
4237                          adapter);
4238         if (rc) {
4239                 dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4240                         vdev->irq, rc);
4241                 goto req_irq_failed;
4242         }
4243
4244         rc = vio_enable_interrupts(vdev);
4245         if (rc) {
4246                 dev_err(dev, "Error %d enabling interrupts\n", rc);
4247                 goto req_irq_failed;
4248         }
4249
4250         crq->cur = 0;
4251         spin_lock_init(&crq->lock);
4252
4253         return retrc;
4254
4255 req_irq_failed:
4256         tasklet_kill(&adapter->tasklet);
4257         do {
4258                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4259         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4260 reg_crq_failed:
4261         dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4262 map_failed:
4263         free_page((unsigned long)crq->msgs);
4264         crq->msgs = NULL;
4265         return retrc;
4266 }
4267
4268 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
4269 {
4270         struct device *dev = &adapter->vdev->dev;
4271         unsigned long timeout = msecs_to_jiffies(30000);
4272         int rc;
4273
4274         if (adapter->resetting && !adapter->wait_for_reset) {
4275                 rc = ibmvnic_reset_crq(adapter);
4276                 if (!rc)
4277                         rc = vio_enable_interrupts(adapter->vdev);
4278         } else {
4279                 rc = init_crq_queue(adapter);
4280         }
4281
4282         if (rc) {
4283                 dev_err(dev, "Couldn't initialize crq. rc=%d\n", rc);
4284                 return rc;
4285         }
4286
4287         adapter->from_passive_init = false;
4288
4289         init_completion(&adapter->init_done);
4290         adapter->init_done_rc = 0;
4291         ibmvnic_send_crq_init(adapter);
4292         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4293                 dev_err(dev, "Initialization sequence timed out\n");
4294                 return -1;
4295         }
4296
4297         if (adapter->init_done_rc) {
4298                 release_crq_queue(adapter);
4299                 return adapter->init_done_rc;
4300         }
4301
4302         if (adapter->from_passive_init) {
4303                 adapter->state = VNIC_OPEN;
4304                 adapter->from_passive_init = false;
4305                 return -1;
4306         }
4307
4308         if (adapter->resetting && !adapter->wait_for_reset)
4309                 rc = reset_sub_crq_queues(adapter);
4310         else
4311                 rc = init_sub_crqs(adapter);
4312         if (rc) {
4313                 dev_err(dev, "Initialization of sub crqs failed\n");
4314                 release_crq_queue(adapter);
4315                 return rc;
4316         }
4317
4318         rc = init_sub_crq_irqs(adapter);
4319         if (rc) {
4320                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4321                 release_crq_queue(adapter);
4322         }
4323
4324         return rc;
4325 }
4326
4327 static struct device_attribute dev_attr_failover;
4328
4329 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
4330 {
4331         struct ibmvnic_adapter *adapter;
4332         struct net_device *netdev;
4333         unsigned char *mac_addr_p;
4334         int rc;
4335
4336         dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
4337                 dev->unit_address);
4338
4339         mac_addr_p = (unsigned char *)vio_get_attribute(dev,
4340                                                         VETH_MAC_ADDR, NULL);
4341         if (!mac_addr_p) {
4342                 dev_err(&dev->dev,
4343                         "(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
4344                         __FILE__, __LINE__);
4345                 return 0;
4346         }
4347
4348         netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
4349                                    IBMVNIC_MAX_QUEUES);
4350         if (!netdev)
4351                 return -ENOMEM;
4352
4353         adapter = netdev_priv(netdev);
4354         adapter->state = VNIC_PROBING;
4355         dev_set_drvdata(&dev->dev, netdev);
4356         adapter->vdev = dev;
4357         adapter->netdev = netdev;
4358
4359         ether_addr_copy(adapter->mac_addr, mac_addr_p);
4360         ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
4361         netdev->irq = dev->irq;
4362         netdev->netdev_ops = &ibmvnic_netdev_ops;
4363         netdev->ethtool_ops = &ibmvnic_ethtool_ops;
4364         SET_NETDEV_DEV(netdev, &dev->dev);
4365
4366         spin_lock_init(&adapter->stats_lock);
4367
4368         INIT_LIST_HEAD(&adapter->errors);
4369         spin_lock_init(&adapter->error_list_lock);
4370
4371         INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
4372         INIT_LIST_HEAD(&adapter->rwi_list);
4373         mutex_init(&adapter->reset_lock);
4374         mutex_init(&adapter->rwi_lock);
4375         adapter->resetting = false;
4376
4377         adapter->mac_change_pending = false;
4378
4379         do {
4380                 rc = ibmvnic_init(adapter);
4381                 if (rc && rc != EAGAIN)
4382                         goto ibmvnic_init_fail;
4383         } while (rc == EAGAIN);
4384
4385         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4386         netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4387         netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4388
4389         rc = device_create_file(&dev->dev, &dev_attr_failover);
4390         if (rc)
4391                 goto ibmvnic_init_fail;
4392
4393         netif_carrier_off(netdev);
4394         rc = register_netdev(netdev);
4395         if (rc) {
4396                 dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
4397                 goto ibmvnic_register_fail;
4398         }
4399         dev_info(&dev->dev, "ibmvnic registered\n");
4400
4401         adapter->state = VNIC_PROBED;
4402
4403         adapter->wait_for_reset = false;
4404
4405         return 0;
4406
4407 ibmvnic_register_fail:
4408         device_remove_file(&dev->dev, &dev_attr_failover);
4409
4410 ibmvnic_init_fail:
4411         release_sub_crqs(adapter);
4412         release_crq_queue(adapter);
4413         free_netdev(netdev);
4414
4415         return rc;
4416 }
4417
4418 static int ibmvnic_remove(struct vio_dev *dev)
4419 {
4420         struct net_device *netdev = dev_get_drvdata(&dev->dev);
4421         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4422
4423         adapter->state = VNIC_REMOVING;
4424         unregister_netdev(netdev);
4425         mutex_lock(&adapter->reset_lock);
4426
4427         release_resources(adapter);
4428         release_sub_crqs(adapter);
4429         release_crq_queue(adapter);
4430
4431         adapter->state = VNIC_REMOVED;
4432
4433         mutex_unlock(&adapter->reset_lock);
4434         device_remove_file(&dev->dev, &dev_attr_failover);
4435         free_netdev(netdev);
4436         dev_set_drvdata(&dev->dev, NULL);
4437
4438         return 0;
4439 }
4440
4441 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
4442                               const char *buf, size_t count)
4443 {
4444         struct net_device *netdev = dev_get_drvdata(dev);
4445         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4446         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
4447         __be64 session_token;
4448         long rc;
4449
4450         if (!sysfs_streq(buf, "1"))
4451                 return -EINVAL;
4452
4453         rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
4454                          H_GET_SESSION_TOKEN, 0, 0, 0);
4455         if (rc) {
4456                 netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
4457                            rc);
4458                 return -EINVAL;
4459         }
4460
4461         session_token = (__be64)retbuf[0];
4462         netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
4463                    be64_to_cpu(session_token));
4464         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
4465                                 H_SESSION_ERR_DETECTED, session_token, 0, 0);
4466         if (rc) {
4467                 netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
4468                            rc);
4469                 return -EINVAL;
4470         }
4471
4472         return count;
4473 }
4474
4475 static DEVICE_ATTR(failover, 0200, NULL, failover_store);
4476
4477 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
4478 {
4479         struct net_device *netdev = dev_get_drvdata(&vdev->dev);
4480         struct ibmvnic_adapter *adapter;
4481         struct iommu_table *tbl;
4482         unsigned long ret = 0;
4483         int i;
4484
4485         tbl = get_iommu_table_base(&vdev->dev);
4486
4487         /* netdev inits at probe time along with the structures we need below*/
4488         if (!netdev)
4489                 return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
4490
4491         adapter = netdev_priv(netdev);
4492
4493         ret += PAGE_SIZE; /* the crq message queue */
4494         ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
4495
4496         for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
4497                 ret += 4 * PAGE_SIZE; /* the scrq message queue */
4498
4499         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4500              i++)
4501                 ret += adapter->rx_pool[i].size *
4502                     IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
4503
4504         return ret;
4505 }
4506
4507 static int ibmvnic_resume(struct device *dev)
4508 {
4509         struct net_device *netdev = dev_get_drvdata(dev);
4510         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4511
4512         if (adapter->state != VNIC_OPEN)
4513                 return 0;
4514
4515         tasklet_schedule(&adapter->tasklet);
4516
4517         return 0;
4518 }
4519
4520 static const struct vio_device_id ibmvnic_device_table[] = {
4521         {"network", "IBM,vnic"},
4522         {"", "" }
4523 };
4524 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
4525
4526 static const struct dev_pm_ops ibmvnic_pm_ops = {
4527         .resume = ibmvnic_resume
4528 };
4529
4530 static struct vio_driver ibmvnic_driver = {
4531         .id_table       = ibmvnic_device_table,
4532         .probe          = ibmvnic_probe,
4533         .remove         = ibmvnic_remove,
4534         .get_desired_dma = ibmvnic_get_desired_dma,
4535         .name           = ibmvnic_driver_name,
4536         .pm             = &ibmvnic_pm_ops,
4537 };
4538
4539 /* module functions */
4540 static int __init ibmvnic_module_init(void)
4541 {
4542         pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
4543                 IBMVNIC_DRIVER_VERSION);
4544
4545         return vio_register_driver(&ibmvnic_driver);
4546 }
4547
4548 static void __exit ibmvnic_module_exit(void)
4549 {
4550         vio_unregister_driver(&ibmvnic_driver);
4551 }
4552
4553 module_init(ibmvnic_module_init);
4554 module_exit(ibmvnic_module_exit);